What are those 6 possible transitions? I get 13 to 15
*P.BeC, P.BW-, P.Da-, *P.Ri, *P.Ya, M4
Every element that is present in E2 but not in E1 will cause a transition. An element that is present in E1 and E2 will not cause a transition. An element that is not present in E1 and E2 will break the pattern at E1.
So again my transition chart from Message 192 modified so that the transition column is either "yes" if it forces a change from E1 to E2 and "no" if it doesn't, where either and "end" or "begin" transition occurs if an element is in E1 but not in E2 OR it is is E2 and not in E1 and "no" if it is the same for E1 and E2:
E1
transition
E2
*P.Al&‘
yes
no
yes
*P.Al&+
no
no
no
*P.Al&-
yes
end
no
*P.BeC&‘
no
begin
yes
*P.BeC&+
no
no
no
*P.BeC&-
no
no
no
*P.BW&‘
yes
end
no
*P.BW&+
yes
end
no
*P.BW&-
no
begin
yes
*P.Da&‘
yes
end
no
*P.Da&+
no
no
no
*P.Da&-
no
begin
yes
*P.En&‘
no
no
no
*P.En&+
no
no
no
*P.En&-
no
no
no
*P.LF&‘
yes
no
yes
*P.LF&+
no
no
no
*P.LF&-
no
no
no
*P.Pi&‘
yes
end
no
*P.Pi&+
no
no
no
*P.Pi&-
no
no
no
*P.Ri&‘
no
begin
yes
*P.Ri&+
no
no
no
*P.Ri&-
no
no
no
*P.Tr&‘
yes
end
no
*P.Tr&+
yes
end
no
*P.Tr&-
no
no
no
*P.WeC&‘
yes
no
yes
*P.WeC&+
no
no
no
*P.WeC&-
yes
end
no
*P.Wo&‘
yes
end
no
*P.Wo&+
no
no
no
*P.Wo&-
no
no
no
*P.WSA&‘
yes
end
no
*P.WSA&+
no
no
no
*P.WSA&-
no
no
no
*P.Ya&‘
no
begin
yes
*P.Ya&+
no
no
no
*P.Ya&-
no
no
no
M1
yes
yes
no
M2
yes
yes
no
M3
no
no
no
M4
no
begin
yes
M5
yes
no
yes
M6
yes
no
no
M7
yes
end
no
M10
no
no
no
M11
no
no
no
M12
no
no
no
M13
yes
end
no
M14
no
no
no
MSE
yes
no
yes
So I get your 6 "begin" transitions but it seems you missed the 14 "end" transitions -- elements that must exit before Event #2 can begin (and before any of the "begin" transitions occur).
And you can also have E1 → E2 without any transition if they have just the people common to both events. This ambiguity would end with a transition to an event element not in both E1 and E2.
RAZD writes:
Caveat: +/- cannot be observed without * appearance of individual.
It can. For Example:
E5: *P.Pi /E6: P.Pi- /E7: P.Pi+, *P.Pi /E8: P.Pi-
Yes, they are identified but it is not an appearance. An appearance for this pattern is defined as someone being named, start to speak or appear. And "he" or "she" is not a name.
Pedantic hair splitting -- you know who is identified by "he" or "she" as plainly as if they were named because you assign +/- to them.
You try to rewrite the pattern rules. For example that every affected person includes an appearance of the person. If you look at table 4 on page 5, then you will see there a lot - or + without *. If they are all replaced from "+" and "-" to "*, +" and "*, -", then an reappearance of this person would not cause a transition.
Except segregating *P.(A)‘, *P.(A)+, *P.(A)- into distinct elements would mean that the appearance of *P.(A)‘ after *P.(A)+ or *P.(A)- would still be a transition because it is (now) a different element.
Now I note that you previously said that P.(A)+ and P.(A)- were not elements, but you are treating them as elements now. Can you clarify this?
With the actual rules:
E1: *P.Da /E2: P.Da- /E3: *P.Da /E9: *P.WeC /E11: *P.Da /E12: P.Da- /E13: *P.Da
For your rules (probably):
E1: *P.Da /E2: P.Da-, *P.Da, *P.WeC, *P.Da, P.Da-, *P.Da
No, I would get (system A with 52 independent elements):
E1: *P.Da‘ /E2: *P.Da- /E3: *P.Da‘ /E9: *P.WeC‘ /E11: *P.Da‘ /E12: *P.Da- /E13: *P.Da‘
or (system B with 28 independent elements):
E1: *P.Da /E2: M- /E3: *P.Da /E9: *P.WeC /E11: *P.Da /E12: M- /E13: *P.Da
You can test your revised system A and B, but they are completely different to the pattern introduced in the paper. ...
Curiously I see them as being the same pattern. Perhaps there is something missing that is not in your verbal descriptions of how the pattern works, or is applied.
... Didn't you said you wanted to reproduce the work and not to create your own?
Oh I am very interested in reproducing the work, and that includes deriving the pattern rather than just accepting it as written.
it seems to me that I can make B15 be P.Da and then E is unique from A-C:
What you replied here has more to do with transitions than with the discussion on hypothetical patterns ... this is about transitions yes?
No. I will explain it again:
An element that is present at E1 and present at E2 will not cause a transition. An element that is present at E1 and not present at E2 will not cause a transition. An element that is not present at E1 and present at E2 will cause a transition. An element that is not present at E1 and not present at E2 will break the pattern.
This is effectively a change to what is acceptable to be part of E2 ... you are now saying that anything left over from E1 can be in E2 even when it is not listed as a part of E2.
Can these E1 hangabouts remain through E3? all the way to E15?
This is detailed explained on page 6: E11: P.Da appears. *P.Da is not part of E9, but part of E11, therefore E11 is triggered. E12: The crew is mentioned. The crew in ST:TNG consist of 1000 persons. *P.Al is not part of E11, but part of E12, therefore E12 is triggered E13: The first officer is named. *P.Ri is not part of E12, but part of E13, therefore E13 is triggered. E14: P.Pi appears. *P.Pi is not part of E13, but part of E14, therefore E14 is triggered. E15: Commander William Riker is commended to be a highly experienced man. P.Ri+ is not part of E14, but part of E15, therefore E15 is triggered
So each event starts when the first member of E(n+1) that is not a member of E(n) is observed regardless of who is hanging around that is not a member of E(n+1) but can be a member of E(1) to E(n)?
That makes the possibilities even more open than before when you have infinite possibilities due to multiple ad nauseum appearances, while maintaining a finite list of pattern breaking possibilities. In essence you have n/∞ → 0 probability of not fitting the "pattern" ...
RAZD writes:
There is no pattern here that I can see: what should I be seeing that I am missing?
You use a random data source. For the random data test it was not possible to create a distinct pattern that fits with the random data.
No, I used selected elements from your "pattern" description in Message 166 "Elements are observed, either singly or in combinations and all with possible repeated appearances" ...
Here it is again (with brown color removed):
Events
Episode A
Episode B
Episode C
Episode D
Episode E
Event #1
P.Al
P.BW
P.Da
P.LF
P.Tr
Event #2
P.BeC
M5
P.LF
P.Al
P.Ya
Event #3
P.En
P.Pi
P.Ri
P.Tr
P.Wo
Event #4
P.Wo
P.Al
M4
M10
‘
Event #5
P.Da
P.En
P.Wo
P.Pi
‘
Event #6
M1
P.Ri
P.Al
M6
‘
Event #7
P.BW
P.Tr
P.WeC
P.Da
‘
Event #8
M4
P.Wo
P.BW
P.Ri
‘
Event #9
P.Ri
P.BeC
P.En
P.Ya
M2
Event #10
P.WSA
‘
P.Pi
‘
P.Da
Event #11
P.WeC
P.LF
‘
‘
M7
Event #12
P.Tr
P.Ya
P.BeC
P.BW
M10
Event #13
M5
M6
M1
M7
P.LF
Event #14
M2
M3
M13
P.BeC
P.Pi
Event #15
M14
P.Da
M12
P.En
M4
Are these elements or are they not part of the groupings you have for the different events that are allowed to occur "either singly or in combinations and all with possible repeated appearances" ?
RAZD writes:
When you assemble mirrors and bits of colored plastic in a kaleidoscope you can create the appearance of colored patterns, where any one would be highly unlikely to occur, but the pattern is an artifact caused by the mirrors, not the colored plastic bits.
It's not an unlikely pattern that occurred once. It is an unlikely pattern with a high complexity, but it didn't occurred only once, it occurred 45 out of 47 times. It's like you use your kaleidoscope and 45 out of 47 times the same unlikely pattern occurs. And a kaleidoscope is not an evolution-like process.
That depends on how the "pattern" is defined. If I define the "pattern" as a symmetrical arrangement of colored elements then anything the kaleidoscope shows will fit that definition of the "pattern" ...
... and in a similar vein if I now define my hypothetical "pattern" to be composed of the following groups, where the elements are observed, either singly or in combinations and all with possible repeated appearances:
P.Al, P.BW, P.Da, P.LF, P.Tr
P.BeC, M5, P.LF, P.Al, P.Ya
P.En, P.Pi, P.Ri, P.Tr, P.Wo
P.Wo, P.Al, M4, M10, ‘
P.Da, P.En, P.Wo, P.Pi, ‘
M1, P.Ri, P.Al, M6, ‘
P.BW, P.Tr, P.WeC, P.Da, ‘
M4, P.Wo, P.BW, P.Ri, ‘
P.Ri, P.BeC, P.En, P.Ya, M2
P.WSA, ‘, P.Pi, ‘, P.Da
P.WeC, P.LF, ‘, ‘, M7
P.Tr, P.Ya, P.BeC, P.BW, M10
M5, M6, M1, M7, P.LF
M2, M3, M13, P.BeC, P.Pi
M14, P.Da, M12, P.En, M4
Now suddenly, where I had 5 episodes that did not show a discernible "pattern" there is a completely perfect fit for each of these hypothetical episodes to the "pattern"
The different kaleidoscope patterns have been fit into a pattern defined to accommodate them.
In fact the "pattern" can fit (1x5!)+ (5x4!)+ (10x3!)+ (10x2!)+ (5x1!) = 325 potential different and unique arrangements of these elements for each event level, or 15 x 325 = 4875 hypothetical episodes from a pattern based on 5 episodes
If I start with one bead in the kaleidoscope I will get the same simple pattern, but as I add beads it becomes more complex, ... if I have one bead of each color -- red, green, blue, yellow, black and white beads, and I define the "pattern" to include at least one appearance of each of these colors then no matter how you twist and turn the kaleidoscope you will get a result that amazingly matches the pattern!
om mani padmi om
The pattern was created for the first 76 episode. It was revised until it did fit with the first 76 episode. ...
Just as my hypothetical "pattern" was based on 5 hypothetical episodes but allows 4875 different hypothetical episodes to fit.
The question that you haven't answered yet is why I should make a pattern this way.
Arguing that I am doing things differently does not refute my approach if you have no reason to take your approach over mine.
RAZD writes:
So I get your 6 "begin" transitions but it seems you missed the 14 "end" transitions -- elements that must exit before Event #2 can begin (and before any of the "begin" transitions occur).
There are no "end transitions" with appearances. Only appearances:* and affected person:+/- were quantised. It would be an other pattern, if they are added.
So the definition in Message 166 "Elements are observed, either singly or in combinations and all with possible repeated appearances" is not complete, each event can also include elements from the previous event.
Curiously I note 2 things: first that the "allowed" list rapidly approaches the full cast, making invalid appearances in later events more and more unlikely, and second that the number of possible hypothetical episodes that would now fit this "pattern" explodes.
I added recurring appearances that are not persons as M's. There are possibly more M's that also fit with pattern but wasn't added yet. On page 15 a possible M15 is mentioned that could appear at E16. In five episodes followed right after E15 speechlessness as a possible M15. It didn't appeared at E1-E15 for the whole data source. I assume that there are not only 15 events. There are possibly a lot more events like 20, 30 or more. The residual uncertainty would drastically increase then. But 1:10^7 is also a good residual uncertainty which is accepted as a declaration of a discovery in science.
So I take hypothetical episode (A) and apply it to episode (B), but it doesn't fit, so I add episode (B) elements to episode (A) elements to make a revised pattern AB and test it on episode (C), but it doesn't fit, so I add episode (C) elements to episodes (A) & (B) elements to make a revised pattern ABC and test it on episode (D), but it doesn't fit, so I add episode (D) elements to episodes (A), (B) & (C) elements to make a revised pattern ABCD and test it on episode (E), but it doesn't fit, so I add episode (E) elements to episodes (A), (B), (C) & (D) elements to make a revised pattern ABCDE ... and each time I increase the pattern elements the number of possible hypothetical episodes that would fit the pattern explodes.
This is making a pattern definition so broad that more and more episodes would fit.
This invalidates any probability calculations you make because you keep expanding the set of possibilities.
... The transitions themselves are also not correctly applied yet.
Then it looks like you've got some explaining to do, Lucy ... and I suggest you go back to the "pattern" description in Message 166 and add these transitions (and who\what can stay from the previous event/s and who\what would be an invalidating elements) between each event:
As before only optional events 10 and/or 11 and/or group {4,5,6,7,8} can be omitted, the order of events cannot be changed.
... That's what the ID proponents claim, that the origin and evolution of life needs an additional information creating process because they can't believe it happened naturally. I wasn't concerned long enough with the origin and evolution of life to be able to tell if this claims have any substance.
They don't, (a) because "information" is either something that evolves naturally or it is unimportant to evolution (based on empirical evidence), and (b) their assembly model is overly simplistic and ignores how molecules and life forms behave.
You should also not attempt to discuss evolution if you do not know how it works.
But only for the first 76 episodes. As stated on page 6: "The pattern was created to fit with season 1, 3 and 4 at the actual start of the episode (00:00). Afterwards it was tested on season 5 and 6 and a random data source". You will find the origin of the pattern in table 5 on page 8. You will see there that this was done only for the first part of the data source. For the second part, season 5 and 6, all positives and negatives are listed in table 5. Three elements didn't fit with the pattern created form the first part of the data source and they are listed in table 5.
It was looked at the complete data source. Only season 2 was skipped because of an other main cast. Season 7 was not looked at, because a residual uncertainty of 1:10^7 was already reached. But if it reassures you, the pattern was also tested for season 7 some time ago and it did fit again 22 out of 23 times.
Curiously that is about what I would expect from the way your pattern is constructed.
The opening scenes have less constraints than the rest of the episode. ...
Opening scenes will always set up a conflict and never resolve one -- because that is how you tell stories.
The pattern is used on the quantisations, not on the original visual information.
For Example:
P.Wo, P.Ya, P.WeC walk onscreen P.Da walks onscreen P.Pi, P.Ri, P.WeC have a conversation
P.Da remains visible in the original visual information, but in the quantisation he doesn't appear again. The pattern is used on the quantisations, not on the original visual information.
So in one episode he stays and in another he doesn't and you think these are the same.
Then I will specify again: Any nontrivial pattern. There are a lot patterns like this:
There is always a person appearing There is always someone affected There is always someone talking
A nontrivial pattern is a pattern far beyond this.
Or the trivial data, if quantised, would be pattern breaking. It seems to me that between this and not counting exits, you are missing a lot of what was going on, with the result that the pattern is trivial.
There are infinite possibilities to fit and not fit with the pattern. For Example:
No, there are a finite number of ways to break the pattern. Think of it this way:
starting with the finite number of patterns that is based on single appearances of an element, and
we would have a finite number of fits and a finite number of fails
to the ones that fit we add elements one at a time with one trial for each different element
again there is a finite number of fits and fails for each round
the proportion of fits to fails would be the same for each round
therefore you would have a limit condition that would approach that proportion
A possible transition is every element that is not part of the current event but part of a next event. An element that is part of the current event will not cause a transition. An element that is not part of the current event or a next event will invalidate the pattern.
There is nothing like "can stay" or "can not stay". The pattern is not used on the original visual information, it is used on the quantisations.
The other transitions and invalidations can be found with the already explained rules too.
Because only *one* invalidating element needs to occur only *once* there are a finite number of deal breakers: repeated or multiple breakers will not make the episode more or less broken.
E1 has 26 deal breakers and E2 has 28
E1 and E2 have infinite possibilities due to repetitions and multiple interactions being allowed The probability of bad fit on E1 is 26/∞ &riarr; 0 The probability of bad fit on E2 is 28/∞ &riarr; 0
With 52 elements (counting +/- and adding 1 for "something else") you have 26/52 = 0.50 for E1 and 28/52 = 0.54 for E2 probability to fail.
IFF your "pattern" is for single independent (stand alone) elements. If your fail elements depend on the existence of a previous element then the likelihood of them occurring is reduced by the probability of the previous element occurring.
ie -- you can't have a P.(A)+ or a P.(A)- unless it is preceded by a *P.(A): that's two possible fail elements dependent on one predecessor.
Looking at your list for E1:
Message 166: Event #1:Elements are observed, either singly or in combinations and all with possible repeated appearances -- P.Al, P.BW, P.Da, P.LF, P.Pi,P.Tr, P.WeC, P.Wo, P.WSA, M1, M2, M5, M6, M7, M13, P.Al-, P.BW+, P.Tr+, P.WeC-.
P.Ya+ and P.Ya- cannot occur because *P.Ya does not exist in E1 and the appearance of P.Ya means you are in E2. Likewise P.BeC+ and P.BeC- cannot occur without *P.BeC causing a transition to E2, *P.En is not listed in E1 or E2 or in your invalidating list, so P.En+ and P.En- also cannot occur without *P.En (which presumably would be an invalidating element), and similarly P.Ri+ and P.Ri- need to be preceded by *P.Ri (which presumably would be an invalidating element), and so we see that the actual possible invalidation elements are reduced to 20 elements.
I meant you didn't retrieved your quantisations from an actual audible and visual data source. If you want to reproduce the work, then you must retrieve your quantisations from an evolution-like process. Arbitrary arranged appearances could contain any arbitrary pattern.
They are take from your quantisations from actual audible and visual data.
And you still haven't said why your elements are grouped the way they are when there is no discernible common thread within each group.
This pattern will have a probability of 1 to fit. But the E1-E15 pattern has only a probability of 0.625 to fit. The 0.625 probability resulted from an experimental test and a theoretical calculation in Message 190 resulted in a probability of p<0.711. To create a comparable pattern, you would have to define a pattern that also fits only about every second time.
Your assumptions are incorrect, but I'm not about to discuss that yet -- I need to know why your groups are selected the way they are.
My simplistic example with 5 different episodes show how a pattern derived from force fitting it to the 5 hypothetical episodes makes it vastly more applicable than when derived from a smaller number of episodes. You used 76 episodes ...
quote:Short form Additionally observation M1 open door, colour black/red M2 weapon, What's that? M3 humour, laughing M4 fire M5 water M6 theft, try to get information (example: sensors) M7 drink M10 past M11 unbelievable attainment M12 temporary interruption M13 long time M14 short time, in a hurry, smoke, gas
Curiously something is missing that would de facto cause an event change: a scene change.
please get over it and stop being such a prima donna
... and RAZD just found out 12 hours later that he had to rethink his fit vs fail and that his 26/∞ argument was just wrong. ...
It wasn't a matter of finding out -- seeing as nobody corrected me -- but of rethinking it. Obviously there are problems with being able to calculate accurate probabilities because of the way some elements interconnect and affect whether other can actually occur, something you failed to notice when compiling your list of deal breakers ...
... and I was honest enough to leave the old argument so you could see it and the changes I made: that should make you happy that I respect your argument enough to do that.
... There is also obviously on one familiar with information science. ...
Where is the science?
... The paper was already looked at by people with enough knowledge about the respective sciences. It doesn't matter to me if this paper will never be published too, ...
If you think you are getting rough treatment, I can guarantee you that it is kinder that a full peer review would be.
We could compare what you have done to curve fitting ...
If I have three data points there are several curves that can be fit to the data (circle, parabola, exponential) that will have different predictive power for testing with the next data point. For the sake of argument we will use the polynomial, as it has advantages for simplicity of solutions:
y = A + Bx + Cx^2
And we use the three known points to solve for A, B and C.
With four data points we can still generate a polynomial curve
y = A + Bx + Cx^2 + Dx^3
And we use the four known points to solve for A, B, C and D.
The more data points the more variables you can solve for, and you will always get 100% fit to the known data.
In addition we can be fairly sure that any new data point near an existing point will be relatively near the curves. Less likely are new data points further from any existing points (especially when outside the data set).
However, we would want to find the simplest solution with the best fit within an acceptable margin of error, and for that you would need to use error function modeling. To generate this you can recursively drop one data point, solving the polynomial for the remaining points, and then finding the root mean square error of each of the dropped points:
Erms = (sum(actual - predicted)^2/N)^(1/2)
Which you may recognize as σ ... but the calculation is simplified by only having one dropped data point, so Erms = |(actual - predicted)| ... and the curve with the lowest Erms would be the best fit of the possible solutions.
Then you can recursively drop a second data point and simplify the formula again and repeat the Erms calculations to find the best fit.
The simplist curve with the smallest Erms within the allowable limits would be the more practical (easy to use) solution.
(There is an online discussion of this type of curve fitting starting at pdf page 22 (book page 4) in Pattern Recognition and Machine Learning by Christopher M. Bishop where he discusses fitting the polynomial to a sine curve.)
My overall impression is that you have not tried to simplify the pattern (equation), preferring to add elements to get a 100% fit, and as a result it is unwieldy and clunky.
You might find that a system with half of your elements would be accurate 95% of the time ... which would still be a strong signal of pattern.
Re: missing element and the problem of stragglers\hangers-on\laggards.
Curiously something is missing that would de facto cause an event change: a scene change.
Did he ever end up explaining how you get from one even to another?
He said his transitions were caused\triggered by the appearance of an element of Event(n+1) that is not in Event(n). And that it doesn't matter if elements in Event(n) stay around into Event(n+1) even if they are not listed as elements for that n+1 event. I have a major problem with this as it changes the definition of Event(n+1) and stops the pattern from being broken from these stragglers\hangers-on\laggards.
Then we have the ongoing problem of arbitrariness, both in how the elements are grouped and in how many events are involved in an episode (15? 14? (minus 10 or 11) 13? (minus 10&11) 10? (minus 4,5,6,7 and 8) or 9 or 8 ... ).
Ask as much as one likes for the rationale behind the choices and there is no answer, just that it IS the pattern.
P.Da remains visible in the original visual information, but in the quantisation he doesn't appear again. The pattern is used on the quantisations, not on the original visual information.
Two things:
First - because the "pattern is used on the quantizations, not on the original visual information" the "pattern" depends on what you chose to quantisize
Second - the "possible fit" is based on how elements are grouped, which appears arbitrary and has not been explained.
RAZD writes:
add these transitions (and who\what can stay from the previous event/s and who\what would be an invalidating elements) between each event:
A possible transition is every element that is not part of the current event but part of a next event. An element that is part of the current event will not cause a transition. An element that is not part of the current event or a next event will invalidate the pattern.
There is nothing like "can stay" or "can not stay". The pattern is not used on the original visual information, it is used on the quantisations.
Now if *P.BeC, *P.Ri, *P.Ya, M4, P.BW-, P.Da- were included in the list of elements for Event #1 then they wouldn't cause a transition to Event #2.
LIkewise if certain elements of Event #1 were moved to Event #2 then their appearance would trigger a transition to Event #2.
Now obviously if you included every possible element into the Event #1 cast there could only be one event, so the question is how and why are the elements divided into different events?
Now I could take the Event #1 cast as a given and then lump everybody in Event #2 (seeing as there are many carry-overs from one event to the next) and then every element not in Event #1 would trigger Event #2.
But why should I take the Event #1 cast as a given - how is it derived?
Well an easy first approximation would be to start with those elements that start the episodes - ie start with the element that is at the start of episode 1 and then add the element that is at the start of episode 2 (if it is not the same element) and so on, until all 76 of the "data set" episodes is included.
With this approach there should be no +/- elements in Event #1 because you can't start with those.
So the Event #1 cast would be defined by episode starters (you could say they trigger the transition to Event #1), and any element not a starter would trigger the transition to Event #Remainder (in this case #2).
Then I could take all the trigger elements - the elements that are not episode starters but which then appear next in each episode and that trigger the transition to the next event ...
So the Event #2 cast would be defined by the Event #2 triggers (they trigger the transition to event #2) ... plus the cast from Event #1 (because I have no reason to remove elements) ... and then any element not an Event #2 cast element would trigger the transition to Event #Remainder (in this case #3).
Likewise the Event #3 cast would be defined by the Event #3 triggers (they trigger the transition to event #3) ... plus the cast from Event #2 (which includes the cast from Event #1 because I still have no reason to remove elements) ... and then any element not an Event #3 cast element would trigger the transition to Event #Remainder (in this case #4).
And we can continue this way until the Event #n cast would include all the elements, there would be none left to cause a transition to a next event.
This pattern would fit the 76 episodes because of the way it is developed. It would also be different from the published one, and likely it would have fewer total events.
This process is dependent on what is chosen to observe as an element that triggers an event: is the picture of a sun (from a planet) different from a picture of a star field? Do I count an empty room (view without people) as an element? Do I then count the room as an element when other elements (people) are included?
Why should I remove any elements of preceding events from later events? For instance I could remove elements in the Event #(n) cast if they did not appear (renew their involvement) in Event #(n-1), but what is the rationale?
So I thought I would play with the pattern discussed in Message 227 and went to look at appendix A thinking I would be able to obtain a sequential listing of the elements used by you to make your pattern and I get this:
Appendix A lists in detail all appearances and persons that are affected by situations and the movement through the pattern for season 1, 3, 4, 5 and 6. Appendix B lists the attempt to apply this pattern at a later starting point. The notation looks like this:
So far so good ... just what I was looking for ...
The caption contains the episode number and the title. The text contains the event numbers, the time the events occur and all appearances. The first three episodes contain more detailed descriptions. For season 3, 4, 5 and 6 there are only appearances and affected persons noted that trigger the next event. Starships at the beginning are also skipped then. (OC) tells the opening credits are included.
Back the bus up, young fella ... where are events 1, 2, 4, 5, 6, 7, 8, and 10? AND you only count elements that trigger the next event in the latter seasons???
So the "pattern" must have been built from the first two seasons in order to know the triggers. Your text talks about adding new elements for later episodes to make them fit the "pattern" ... did you go back and review the first two seasons for these additional elements to see if they changed the documentation for those seasons?
(notes) 9 (1x01 E12) The crew is mentioned, the crew in ST:TNG consist of 1000 persons 10 (1x03 E6, E12, E14) A starfleet vessel is named 11 (1x03 E8) Problems on the starfleet science vessel are mentioned 12 (1x03 E9) Number of persons visible at the same time exceeds 5 13 (1x03 E11) Data starts to speak 14 (1x03 E13) Woman without special abilities, appears in: 1x03, 3x06, 3x22, 3x25 and 5x06 15 (1x03 E13) Riker appears again through disappearance before, Picard was visible since E7 without interruption, he does not appear again
Now I notice that disappearances are mentioned even though you tell me they are not part of the "pattern" ...
Presumably there is no documentation for episode 2 because it is a continuation ... even though there would be a recap introduction ...
Seems rather haphazard recording of data if you don't have elements listed for some of the events but then claim that the events occur as documented in the pattern.
This reviewer throws the paper back on your desk and says "do over" -- add a section to explain how the pattern is developed (so that it can be reproduced), how the events are determined (so that they can be reproduced), and reasons for grouping elements into the events (so that they can be reproduced), and provide all the data in chronological sequences, without listing events (events are not part of the data, they are what develops from the data as you develop the pattern.
P.Ya+ and P.Ya- cannot occur because *P.Ya does not exist in E1 and the appearance of P.Ya means you are in E2. Likewise P.BeC+ and P.BeC- cannot occur without *P.BeC causing a transition to E2, *P.En is not listed in E1 or E2 or in your invalidating list, so P.En+ and P.En- also cannot occur without *P.En (which presumably would be an invalidating element), and similarly P.Ri+ and P.Ri- need to be preceded by *P.Ri (which presumably would be an invalidating element), and so we see that the actual possible invalidation elements are reduced to 20 elements.
You are actually right about this. But that's only for E1. The later events have a possible appearance before.
Thanks . Good to know you are not one of those people that cannot admit to errors.
But I have a feeling there will be cases in other episodes where *P.(A) has not appeared but P.(A)+/- could be listed as deal breakers if we are not careful.
RAZD writes:
The pattern is used on the quantisations, not on the original visual information. ... A possible fit is:
P.Da remains visible in the original visual information, but in the quantisation he doesn't appear again. The pattern is used on the quantisations, not on the original visual information.
So in one episode he stays and in another he doesn't and you think these are the same.
Not really. I'm just saying that there is a difference to people remaining and people leaving a scene, and that this muddles the dependent P.(A)+/- assignments.
RAZD writes:
And you still haven't said why your elements are grouped the way they are when there is no discernible common thread within each group.
The pattern was created to fit with the first 76 episodes. If you remove *P.Da from E7, then the pattern wouldn't fit anymore with 1x03 (table 5 page 8). If you add *P.Da to E6, then it wouldn't fit anymore with 1x03 too. The pattern is the simplest pattern known to me, that fits with this whole first part and can describe all *, + and - over a time of nearly a few minutes.
That still does not explain the method used to determine the event groups (or how events are defined) and this makes it difficult to duplicate.
RAZD writes:
Well an easy first approximation would be to start with those elements that start the episodes - ie start with the element that is at the start of episode 1 and then add the element that is at the start of episode 2 (if it is not the same element) and so on, until all 76 of the "data set" episodes is included.
That's how it was done. There are 76 episode. The pattern is the simplest pattern known to me, that fits with this whole first part and can describe all *, + and - over a time of nearly a few minutes.
That's a start, but as noted, I ran into a couple of problems doing that.
RAZD writes:
And we can continue this way until the Event #n cast would include all the elements, there would be none left to cause a transition to a next event.
True. It wouldn't describe the happenings over time anymore.
The way I saw it was that the probability of a deal breaker decreased with each event if you don't have a mechanism (and a reason for it) for removing elements from the event cast, so you have increasing probability of pattern fit when it should be constant or increasing.
RAZD writes:
This process is dependent on what is chosen to observe as an element that triggers an event: is the picture of a sun (from a planet) different from a picture of a star field? Do I count an empty room (view without people) as an element? Do I then count the room as an element when other elements (people) are included?
Only *, + and - for persons were observed first. All M's that appeared always at the same events were then added to the pattern. There are possibly more M's that also always appear at the same events.
Curiously this makes the pattern less restrictive rather than more ... adding optional alternatives when you don't need it to define the pattern.
RAZD writes:
Back the bus up, young fella ... where are events 1, 2, 4, 5, 6, 7, 8, and 10? AND you only count elements that trigger the next event in the latter seasons???
From page 6 and [Msg=190]: "At the events 1, 3, 4 and 5 the pattern is allowed to start". For 1x01 the pattern starts at E3. The optional events 10 and/or 11 and/or group {4,5,6,7,8} can be omitted. E{4,5,6,7,8} and E10 were omitted.
More options ... so NOW we have (updating the pattern in Message 166:
When you said:
Only optional events 10 and/or 11 and/or group {4,5,6,7,8} can be omitted, the order of events cannot be changed.
We now have optional events 1, 2 and 3 ...
So the pattern has ...
four variations,
one with E1 (and E2 and E3), and
one without E1, which then has three sub-variations,
one with E2 (and E3), and
one without E2, which then has two sub-sub-variations,
one with E3 and
one without,
two variation, one with E4-E8 and one without, and both variations have
two variation, one with E10 and one without, and both variations have
two variation, one with E11 and one without, ...
... for a total of 4x2x2x2 different pattern variations possible or 32 different patterns ...
And now when we go back to the deal breakers we have several possibilities where instead of (P.En) invalidating Event #1 it could just be due to E1 and E2 being skipped (depending on who else is involved) -- the invalidations would have to work in ALL 32 variations.
So what I would rather see is that the pattern is fixed to 15 events and ones that have missing events -- but the rest of the sequence holds -- would be places where the pattern misses the predicted values
ie -- Events 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 and 15 would have 14/15 accuracy (93%) and 1, 2, 3, 9, 10, 11, 12, 13, 14 and 15 would have 10/15 accuracy (67%)
And really, the only thing that invalidates your "pattern" are "events" that are out of sequence rather than elements of patterns -- because they do show up in other events.
RAZD writes:
But if it reassures you, the pattern was also tested for season 7 some time ago and it did fit again 22 out of 23 times.
Curiously that is about what I would expect from the way your pattern is constructed.
I assume you mean that it would always fit with every data source? The probability was tested to 0.625 to fit with random data and calculated to <0.711 in [Msg=190]. If you are agreeing with this calculation and the test, then you would expect only a 14 out of 23 fit.
Curiously I think your calculations are flawed by not properly accounting for the multiple (32) patterns within your overall pattern, any one of which can be fit by the new season episodes.
Remember when I said I felt you should reduce the number of variables and be happy with a result that was ~95% accurate?
This is usually regarded in science as a good match between model and reality, and certainly good enough to produce predictions in most cases.
The problem with having 24 optional patterns is that it is difficult to use for predictions -- you would have 24 different predictions ... which each would need to be tested.
RAZD writes:
So the pattern has ...
four variations,
one with E1 (and E2 and E3), and
one without E1, which then has three sub-variations,
one with E2 (and E3), and
one without E2, which then has two sub-sub-variations,
one with E3 and
one without,
E2 is not a possible start. The pattern has
four variations,
one with E1 as start, which then has 8 sub-variations,
E4-E8:y; E10:y; E11:y
E4-E8:y; E10:y; E11:n
E4-E8:y; E10:n; E11:y
E4-E8:y; E10:n; E11:n
E4-E8:n; E10:y; E11:y
E4-E8:n; E10:y; E11:n
E4-E8:n; E10:n; E11:y
E4-E8:n; E10:n; E11:n
one with E3 as start, which then has 8 sub-variations,
E4-E8:y; E10:y; E11:y
E4-E8:y; E10:y; E11:n
E4-E8:y; E10:n; E11:y
E4-E8:y; E10:n; E11:n
E4-E8:n; E10:y; E11:y
E4-E8:n; E10:y; E11:n
E4-E8:n; E10:n; E11:y
E4-E8:n; E10:n; E11:n
one with E4 as start, which then has 4 sub-variations,
E10:y; E11:y
E10:y; E11:n
E10:n; E11:y
E10:n; E11:n
one with E5 as start, which then has 4 sub-variations,
E10:y; E11:y
E10:y; E11:n
E10:n; E11:y
E10:n; E11:n
for a total of 24 different pattern variations. Only E1, E3, E4 or E5 can be arbitrary chosen. Which events are omitted is decided by the elements that occur. If at E3 an element of E9 occurs which is not part of E3, then E4-E8 are skipped.
Or we count it as points where the model with 15 events missed predicting the data:
four cases,
one with E1 as start, which then has 8 sub-variations,
E4-E8:y; E10:y; E11:y -- all events from E1 to E15 modeled: 100.0% fit
E4-E8:y; E10:y; E11:n -- one event missed, 14 of 15 events modeled: 93.3% fit
E4-E8:y; E10:n; E11:y -- one event missed, 14 of 15 events modeled: 93.3% fit
E4-E8:y; E10:n; E11:n -- two events missed, 13 of 15 events modeled: 86.7% fit
E4-E8:n; E10:y; E11:y -- five events missed, 10 of 15 events modeled: 66.7% fit
E4-E8:n; E10:y; E11:n -- six events missed, 9 of 15 events modeled: 60.0% fit
E4-E8:n; E10:n; E11:y -- six events missed, 9 of 15 events modeled: 60.0% fit
E4-E8:n; E10:n; E11:n -- seven events missed, 8 of 15 events modeled: 53.3% fit
one with E3 as start, which then has 8 sub-variations,
E4-E8:y; E10:y; E11:y -- two events missed, 13 of 15 events modeled: 86.7% fit
E4-E8:y; E10:y; E11:n -- three events missed, 12 of 15 events modeled: 80.0% fit
E4-E8:y; E10:n; E11:y -- three events missed, 12 of 15 events modeled: 80.0% fit
E4-E8:y; E10:n; E11:n -- four events missed, 11 of 15 events modeled: 73.3% fit
E4-E8:n; E10:y; E11:y -- seven events missed, 8 of 15 events modeled: 53.3% fit
E4-E8:n; E10:y; E11:n -- eight events missed, 7 of 15 events modeled: 46.7% fit
E4-E8:n; E10:n; E11:y -- eight events missed, 7 of 15 events modeled: 46.7% fit
E4-E8:n; E10:n; E11:n -- nine events missed, 6 of 15 events modeled: 40.0% fit
one with E4 as start, which then has 4 sub-variations,
E10:y; E11:y -- three events missed, 12 of 15 events modeled: 80.0% fit
E10:y; E11:n -- four events missed, 11 of 15 events modeled: 73.3% fit
E10:n; E11:y -- four events missed, 11 of 15 events modeled: 73.3% fit
E10:n; E11:n -- five events missed, 10 of 15 events modeled: 66.7% fit
one with E5 as start, which then has 4 sub-variations,
E10:y; E11:y -- four events missed, 11 of 15 events modeled: 73.3% fit
E10:y; E11:n -- five events missed, 10 of 15 events modeled: 66.7% fit
E10:n; E11:y -- five events missed, 10 of 15 events modeled: 66.7% fit
E10:n; E11:n -- six events missed, 9 of 15 events modeled: 60.0% fit
Now because the model is biased to having 15 events all the errors are to the side of missed events with no inserted events ... except for invalid transitions ...
On the other hand, if you start the pattern at E3 and delete E11 and E12, 11 fit points:
Remember when I said I felt you should reduce the number of variables and be happy with a result that was ~95% accurate?
This is usually regarded in science as a good match between model and reality, and certainly good enough to produce predictions in most cases.
The problem with having 24 optional patterns is that it is difficult to use for predictions -- you would have 24 different predictions ... which each would need to be tested.
RAZD writes:
So the pattern has ...
four variations,
one with E1 (and E2 and E3), and
one without E1, which then has three sub-variations,
one with E2 (and E3), and
one without E2, which then has two sub-sub-variations,
one with E3 and
one without,
E2 is not a possible start. The pattern has
four variations,
one with E1 as start, which then has 8 sub-variations,
E4-E8:y; E10:y; E11:y
E4-E8:y; E10:y; E11:n
E4-E8:y; E10:n; E11:y
E4-E8:y; E10:n; E11:n
E4-E8:n; E10:y; E11:y
E4-E8:n; E10:y; E11:n
E4-E8:n; E10:n; E11:y
E4-E8:n; E10:n; E11:n
one with E3 as start, which then has 8 sub-variations,
E4-E8:y; E10:y; E11:y
E4-E8:y; E10:y; E11:n
E4-E8:y; E10:n; E11:y
E4-E8:y; E10:n; E11:n
E4-E8:n; E10:y; E11:y
E4-E8:n; E10:y; E11:n
E4-E8:n; E10:n; E11:y
E4-E8:n; E10:n; E11:n
one with E4 as start, which then has 4 sub-variations,
E10:y; E11:y
E10:y; E11:n
E10:n; E11:y
E10:n; E11:n
one with E5 as start, which then has 4 sub-variations,
E10:y; E11:y
E10:y; E11:n
E10:n; E11:y
E10:n; E11:n
for a total of 24 different pattern variations. Only E1, E3, E4 or E5 can be arbitrary chosen. Which events are omitted is decided by the elements that occur. If at E3 an element of E9 occurs which is not part of E3, then E4-E8 are skipped.
Or we count it as points where the model with 15 events missed predicting the data:
four cases,
one with E1 as start, which then has 8 sub-variations,
E4-E8:y; E10:y; E11:y -- all events from E1 to E15 modeled: 100.0% fit
E4-E8:y; E10:y; E11:n -- one event missed, 14 of 15 events modeled: 93.3% fit
E4-E8:y; E10:n; E11:y -- one event missed, 14 of 15 events modeled: 93.3% fit
E4-E8:y; E10:n; E11:n -- two events missed, 13 of 15 events modeled: 86.7% fit
E4-E8:n; E10:y; E11:y -- five events missed, 10 of 15 events modeled: 66.7% fit
E4-E8:n; E10:y; E11:n -- six events missed, 9 of 15 events modeled: 60.0% fit
E4-E8:n; E10:n; E11:y -- six events missed, 9 of 15 events modeled: 60.0% fit
E4-E8:n; E10:n; E11:n -- seven events missed, 8 of 15 events modeled: 53.3% fit
one with E3 as start, which then has 8 sub-variations,
E4-E8:y; E10:y; E11:y -- two events missed, 13 of 15 events modeled: 86.7% fit
E4-E8:y; E10:y; E11:n -- three events missed, 12 of 15 events modeled: 80.0% fit
E4-E8:y; E10:n; E11:y -- three events missed, 12 of 15 events modeled: 80.0% fit
E4-E8:y; E10:n; E11:n -- four events missed, 11 of 15 events modeled: 73.3% fit
E4-E8:n; E10:y; E11:y -- seven events missed, 8 of 15 events modeled: 53.3% fit
E4-E8:n; E10:y; E11:n -- eight events missed, 7 of 15 events modeled: 46.7% fit
E4-E8:n; E10:n; E11:y -- eight events missed, 7 of 15 events modeled: 46.7% fit
E4-E8:n; E10:n; E11:n -- nine events missed, 6 of 15 events modeled: 40.0% fit
one with E4 as start, which then has 4 sub-variations,
E10:y; E11:y -- three events missed, 12 of 15 events modeled: 80.0% fit
E10:y; E11:n -- four events missed, 11 of 15 events modeled: 73.3% fit
E10:n; E11:y -- four events missed, 11 of 15 events modeled: 73.3% fit
E10:n; E11:n -- five events missed, 10 of 15 events modeled: 66.7% fit
one with E5 as start, which then has 4 sub-variations,
E10:y; E11:y -- four events missed, 11 of 15 events modeled: 73.3% fit
E10:y; E11:n -- five events missed, 10 of 15 events modeled: 66.7% fit
E10:n; E11:y -- five events missed, 10 of 15 events modeled: 66.7% fit
E10:n; E11:n -- six events missed, 9 of 15 events modeled: 60.0% fit
Now because the model is biased to having 15 events all the errors are to the side of missed events with no inserted events ... except for invalid transitions ...
On the other hand, if you start the pattern at E3 and delete E11 and E12, for 11 fit points ... or start at E4 with 10 fit points:
Case
# events
15E error
15E error^2
11E error
11E error^2
10E error
10E error^2
1
15
+0
0
+4
16
+5
25
2
14
-1
1
+3
9
+4
16
3
14
-1
1
+3
9
+4
16
4
13
-2
4
+2
4
+3
9
5
10
-5
25
-1
1
0
0
6
9
-6
36
-2
4
-1
1
7
9
-6
36
-2
4
-1
1
8
8
-7
49
-3
9
-2
4
9
13
-2
4
+2
4
+3
9
10
12
-3
9
+1
1
+2
4
11
12
-3
9
+1
1
+2
4
12
11
-4
16
0
0
+1
1
13
8
-7
49
-3
9
-2
4
14
7
-8
64
-4
16
-3
9
15
7
-8
64
-4
16
-3
9
16
6
-9
81
-5
25
-4
16
17
12
-3
9
+1
1
+2
4
18
11
-4
16
0
0
+1
1
19
11
-4
16
0
0
+1
1
20
10
-5
25
-1
1
0
0
21
11
-4
16
0
0
+1
1
22
10
-5
25
-1
1
0
0
23
10
-5
25
-1
1
0
0
24
9
-6
36
-2
4
-1
1
sum
-108
616
-12
136
+12
136
average
-4.50
-0.50
0.50
Erms
+/-5.07
+/-2.38
+/-2.38
So both the 10 event model and the 11 event model have higher predictive value than the 15 event model, and the 10 event model has fewer parts, so it is the better model.
These two single pattern models probably account for a good proportion of your multi-pattern model.
My claim, yet to be refuted or even addressed by you is that simply by trying to write a good story, and sticking to well accepted, and well worn conventions, human actions is sufficient to explain the relationships you observed regarding P.Ya. (Well that and your flexible rules about what constitutes an appearance). I suspect that stories that violate your observation would not pass the smell test of being acceptable stories for a largely Christian audience.
Indeed, the feedback from the audience on what are good episodes goes into editors trying to repeat those successful episodes, and the more they can do that then the more successful the series will be, leading to more seasons ...
It's a feedback loop, much like evolution: variation followed by selection followed by another round of variation followed by selection etc etc etc
So no one person need design the "pattern" rather it can easily be an emergent property of the whole process, and author, editor, producer, actor, station, audience all function as an ber entity to cause the "pattern" ...
There are only 4 real different predictions. The sub-variations can't be chosen arbitrary. Only in this cases are more than one sub-variation possible:
E4-E8: M4 appears at E3. M4 is part of E4 and E9, but not part of E3. This has never happened.
That something has not happened does not mean it has to be excluded from your calculations if it is possible under your rules.
What you CAN do is go through the episodes and class them by the 24 possible cases and then calculate the Root Mean Square Error for the whole data set. Here is the table again, with the events listed so you can compare them:
And I have gone through the season 1 episodes in Appendix A and extracted the event chronology per your descriptions there:
Table B
Episode
Events Listed
Case No
1x01 Encounter At Farpoint (1)
E3, E9, E11, E12, E13, E14, E15
15
1x02 Encounter At Farpoint (2)
(Continuation of 1x01)
na
1x03 The Naked Now
E4, E5, E6, E7, E8, E9, E11, E12, E13, E14, E15
17
1x04 Code Of Honor
E3, E9, E11, E12, E13, E14, E15
15
1x05 The Last Outpost
E3, E9, E11, E12, E13, E14, E15
15
1x06 Where No One Has Gone Before
E4, E5, E6, E7, E8, E9, E11, E12, E13, E14, E15
19
1x07 Lonely Among Us
E4, E5, E6, E7, E8, E9, E10, E12, E13, E14, E15
18
1x08 Justice
E3, E9, E11, E12, E13, E14, E15
15
1x09 The Battle
E3, E9, E11, E12, E13, E14, E15
15
1x10 Hide And Q
E1, E2, E3, E9, E12, E13, E14, E15
8
1x11 Haven
E4, E5, E6, E7, E8, E9, E11, E12, E13, E14, E15
19
1x12 The Big Goodbye
E4, E5, E6, E7, E8, E9, E11, E12, E13, E14, E15
19
1x13 Datalore
E3, E9, E11, E12, E13, E14, E15
15
1x14 Angel One
E3, E4, E5, E6, E7, E8, E9, E12, E13, E14, E15
12
1x15 11001001
E3, E9, E11, E12, E13, E14, E15
15
1x16 Too Short A Season
E3, E9, E11, E12, E13, E14, E15
15
1x17 When The Bough Breaks
E5, E6, E7, E8, E9, E11, E12, E13, E14, E15
23
1x18 Home Soil
E3, E9, E11, E12, E13, E14, E15
15
1x19 Coming of Age
E3, E9, E11, E12, E13, E14, E15
15
1x20 Heart Of Glory
E4, E5, E6, E7, E8, E9, E11, E12, E13, E14, E15
19
1x21 The Arsenal Of Freedom
E1, E2, E3, E9, E11, E12, E13, E14, E15
7
1x22 Symbiosis
E4, E5, E6, E7, E8, E9, E11, E12, E13, E14, E15
19
1x23 Skin Of Evil
E3, E9, E11, E12, E13, E14, E15
15
1x24 We'll Always Have Paris
E4, E5, E6, E7, E8, E9, E12, E13, E14, E15
19
1x25 Conspiracy
E3, E9, E11, E12, E13, E14, E15
15
1x26 The Neutral Zone
E4, E5, E6, E7, E8, E9, E11, E12, E13, E14, E15
19
So now we can combine tables A and B to see how well your "pattern" fits the first season:
Table C
Case
Description
num occurs
events per case
15E error
num*15E error^2
11E error
num*11E error^2
10E error
num*10E error^2
1
E1-E15
0
-
-
-
-
-
-
-
2
E1-E10,,E12-E15
0
-
-
-
-
-
-
-
3
E1-E9,,E11-E15
0
-
-
-
-
-
-
-
4
E1-E9,,,E12-E15
0
-
-
-
-
-
-
-
5
E1-E3,,,,,,E9-E15
0
-
-
-
-
-
-
-
6
E1-E3,,,,,,E9,E10,,E12-E15
0
-
-
-
-
-
-
-
7
E1-E3,,,,,,E9,,E11-E15
1
9
-6
36
-2
4
-1
1
8
E1-E3,,,,,,E9,,,E12-E15
1
8
-7
49
-3
9
-2
4
9
,,E3-E15
0
-
-
-
-
-
-
-
10
,,E3-E10,,E12-E15
0
-
-
-
-
-
-
-
11
,,E3-E9,,E11-E15
0
-
-
-
-
-
-
-
12
,,E3-E9,,,E12-E15
1
11
-4
16
0
0
+1
1
13
,,E3,,,,,,E9-E15
0
-
-
-
-
-
-
-
14
,,E3,,,,,,E9,E10-E15
0
-
-
-
-
-
-
-
15
,,E3,,,,,,E9,,E11-E15
12
7
-96
768
-48
192
-36
108
17
,,,E4-E15
1
12
-3
9
+1
1
+2
4
18
,,,E4-E10,,E12-E15
1
11
-4
16
0
0
+1
1
19
,,,E4-E9,,E11-E15
7
11
-28
112
0
0
+7
7
20
,,,E4-E9,,,E12-E15
0
-
-
-
-
-
-
-
21
,,,,E5-E15
0
-
-
-
-
-
-
-
22
,,,,E5-E10,,E12-E15
0
-
-
-
-
-
-
-
23
,,,,E5-E9,,E11-E15
1
10
-5
25
-1
1
0
0
24
,,,,E5-E9,,,E12-E15
0
-
-
-
-
-
-
-
sums
252
-153
1031
-53
207
-28
126
sum/num
-0.61
4.09
-0.21
0.82
-0.11
0.50
Erms
+/-2.02
+/-0.91
+/-0.71
Curiously I note that the most common pattern is Case 15, covering 12 of the 24 episodes, yet it still results in relatively high error counts for the three models here.
And 23 of the 24 episodes did not have E10.
Based on my analysis I would combine E1, E2 and E3 into one event, E4, E5, E6, E7 and E8 into one event and E10 and E11 into one event, which I would designate
S1 (E1, E2 and E3), S2 ( E4, E5, E6, E7 and E8), S3 (E9), S4 (E10 and E11), S5 (E12), S6 (E13), S7 (E14) and S8 (E15) and the pattern would be:
... based on your division of events in the episodes as you have recorded.
Yes, you have accounted 616 errors ...
That was the sum of the errors square so that the Erms could be calculated.
Curiously I find that the similarity of Erms for the different models means that there is a large variation from each one -- and that this is seen with none of them being close to the most common case, case 15. In the revised system with eight sequences (S1 to S8) there would be one error (S4 missing), so I would expect much closer results using these sequences.
Continuing on from the revised pattern to include 8 appearance sequences, the variations would be
S1 (E1, E2 and E3), S2 ( E4, E5, E6, E7 and E8), S3 (E9), S4 (E10 and E11), S5 (E12), S6 (E13), S7 (E14) and S8 (E15) ,
S1 (E1, E2 and E3), S2 ( E4, E5, E6, E7 and E8), S3 (E9), S4 (E10 and E11), S5 (E12), S6 (E13), S7 (E14) and S8 (E15) ,
S1 (E1, E2 and E3), S2 ( E4, E5, E6, E7 and E8), S3 (E9), (E10 and E11), S4, S5 (E12), S6 (E13), S7 (E14) and S8 (E15) ,
S1 (E1, E2 and E3), S2 ( E4, E5, E6, E7 and E8), S3 (E9), S4 (E10 and E11), S5 (E12), S6 (E13), S7 (E14) and S8 (E15) ,
S1 (E1, E2 and E3), S2 ( E4, E5, E6, E7 and E8), S3 (E9), S4 (E10 and E11), S5 (E12), S6 (E13), S7 (E14) and S8 (E15) ,
S1 (E1, E2 and E3), S2 ( E4, E5, E6, E7 and E8), S3 (E9), S4 (E10 and E11), S5 (E12), S6 (E13), S7 (E14) and S8 (E15) ,
S1 (E1, E2 and E3), S2 ( E4, E5, E6, E7 and E8), S3 (E9), (E10 and E11), S4, S5 (E12), S6 (E13), S7 (E14) and S8 (E15) ,
S1 (E1, E2 and E3), S2 ( E4, E5, E6, E7 and E8), S3 (E9), S4 (E10 and E11), S5 (E12), S6 (E13), S7 (E14) and S8 (E15) .
Table A
Sequence No
Description
events per sequence
8S error
8S error^2
7S error
7S error^2
6S error
6S error^2
1
S1,S2,S3,S4,S5,S6,S7,S8
8
+0
0
+1
1
+2
4
2
S1,S2,S3, ,S5,S6,S7,S8
7
-1
0
+0
0
+1
1
3
S1, ,S3,S4,S5,S6,S7,S8
7
-1
0
+0
0
+1
1
4
S1, ,S3, ,S5,S6,S7,S8
6
-2
4
-1
1
+0
0
5
,S2,S3,S4,S5,S6,S7,S8
7
-1
0
+0
0
+1
1
6
,S2,S3, ,S5,S6,S7,S8
6
-2
4
-1
1
+0
0
7
, ,S3,S4,S5,S6,S7,S8
6
-2
4
-1
1
+0
0
8
, ,S3, ,S5,S6,S7,S8
5
-3
9
-2
4
-1
1
sumS
-12
21
-4
8
+4
8
average
-1.50
-0.50
0.50
Erms
+/-1.63
+/-1.0
+/-1.0
And I have gone through the season 1 episodes in Appendix A and extracted the event chronology per your descriptions there:
Table B
Episode
Events Listed
Sequence No
1x01 Encounter At Farpoint (1)
S1, , S3, S4, S5, S6, S7, S8
3
1x02 Encounter At Farpoint (2)
(Continuation of 1x01)
na
1x03 The Naked Now
, S2, S3, S4, S5, S6, S7, S8
5
1x04 Code Of Honor
S1, , S3, S4, S5, S6, S7, S8
3
1x05 The Last Outpost
S1, , S3, S4, S5, S6, S7, S8
3
1x06 Where No One Has Gone Before
, S2, S3, S4, S5, S6, S7, S8
5
1x07 Lonely Among Us
, S2, S3, S4, S5, S6, S7, S8
5
1x08 Justice
S1, , S3, S4, S5, S6, S7, S8
3
1x09 The Battle
S1, , S3, S4, S5, S6, S7, S8
3
1x10 Hide And Q
S1, S3, , S5, S6, S7, S8
4
1x11 Haven
, S2, S3, S4, S5, S6, S7, S8
5
1x12 The Big Goodbye
, S2, S3, S4, S5, S6, S7, S8
5
1x13 Datalore
S1, , S3, S4, S5, S6, S7, S8
3
1x14 Angel One
S1, S2, S3, , S5, S6, S7, S8
2
1x15 11001001
S1, , S3, S4, S5, S6, S7, S8
3
1x16 Too Short A Season
S1, , S3, S4, S5, S6, S7, S8
3
1x17 When The Bough Breaks
, S2, S3, S4, S5, S6, S7, S8
5
1x18 Home Soil
S1, , S3, S4, S5, S6, S7, S8
3
1x19 Coming of Age
S1, , S3, S4, S5, S6, S7, S8
3
1x20 Heart Of Glory
, S2, S3, S4, S5, S6, S7, S8
5
1x21 The Arsenal Of Freedom
S1, , S3, S4, S5, S6, S7, S8
3
1x22 Symbiosis
, S2, S3, S4, S5, S6, S7, S8
5
1x23 Skin Of Evil
S1, , S3, S4, S5, S6, S7, S8
3
1x24 We'll Always Have Paris
, S2, S3, , S5, S6, S7, S8
6
1x25 Conspiracy
S1, , S3, S4, S5, S6, S7, S8
3
1x26 The Neutral Zone
, S2, S3, S4, S5, S6, S7, S8
5
And as before, accounting for the frequency of sequences in the first season:
Table C
Sequence No
Description
events per sequence
freq, f
8S error
8S error^2
7S error
7S error^2
6S error
6S error^2
Sequence
Description
num
f
err
err^2
err
err^2
err
err^2
1
S1,S2,S3,S4,S5,S6,S7,S8
8
0
0
0
0
0
0
0
2
S1,S2,S3,,S5,S6,S7,S8
7
1
-1
1
0
0
1
1
3
S1,,S3,S4,S5,S6,S7,S8
7
13
-13
13
0
0
13
13
4
S1,,S3,,S5,S6,S7,S8
6
1
-2
4
-1
1
0
0
5
,S2,S3,S4,S5,S6,S7,S8
7
9
-9
9
0
0
9
9
6
,S2,S3,,S5,S6,S7,S8
6
1
-2
4
-1
1
0
0
7
,,S3,S4,S5,S6,S7,S8
6
0
0
0
0
0
0
0
8
,,S3,,S5,S6,S7,S8
5
0
0
0
0
0
0
0
SUMS
52
25
-27
31
-2
2
23
23
SUMS/num
1
-1.08
-0.08
0.92
Erms
1.11
0.28
0.96
This tells us that one of the 7 sequence patterns is better than either the 8 sequence or the 6 sequence patterns and that it is better that the event patterns, but it is still a simplistic analysis: in reality two different 7 sequence patterns differ from each other at two points, not 0, and which one is better at matching the episode data would take another analysis with absolute values for departures, not +/- values.
That is best done with a modified table B to compare each episode with the three 7 sequence patterns SN2, SN3 and SN5:
Table D
Episode
Description
Sequence No's
|err|: SQN2
|err|: SQN2^2
|err|: SQN3
|err|: SQN3^2
|err|: SN5
|err|: SQN5^2
1x01 Encounter At Farpoint (1)
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x02 Encounter At Farpoint (2)
(Continuation of 1x01)
1x03 The Naked Now
, S2, S3, S4, S5, S6, S7, S8
5
2
4
2
4
0
0
1x04 Code Of Honor
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x05 The Last Outpost
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x06 Where No One Has Gone Before
, S2, S3, S4, S5, S6, S7, S8
5
2
4
2
4
0
0
1x07 Lonely Among Us
, S2, S3, S4, S5, S6, S7, S8
5
2
4
2
4
0
0
1x08 Justice
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x09 The Battle
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x10 Hide And Q
S1, , S3, , S5, S6, S7, S8
4
1
1
1
1
3
9
1x11 Haven
, S2, S3, S4, S5, S6, S7, S8
5
2
4
2
4
0
0
1x12 The Big Goodbye
, S2, S3, S4, S5, S6, S7, S8
5
2
4
2
4
0
0
1x13 Datalore
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x14 Angel One
S1, S2, S3, , S5, S6, S7, S8
2
0
0
2
4
0
0
1x15 11001001
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x16 Too Short A Season
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x17 When The Bough Breaks
, S2, S3, S4, S5, S6, S7, S8
5
2
4
2
4
0
0
1x18 Home Soil
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x19 Coming of Age
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x20 Heart Of Glory
, S2, S3, S4, S5, S6, S7, S8
5
2
4
2
4
0
0
1x21 The Arsenal Of Freedom
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x22 Symbiosis
, S2, S3, S4, S5, S6, S7, S8
5
2
4
2
4
0
0
1x23 Skin Of Evil
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x24 We'll Always Have Paris
, S2, S3, , S5, S6, S7, S8
6
1
1
3
9
1
1
1x25 Conspiracy
S1, , S3, S4, S5, S6, S7, S8
3
2
4
0
0
2
4
1x26 The Neutral Zone
, S2, S3, S4, S5, S6, S7, S8
5
2
4
2
4
0
0
sums
25
46
90
24
50
30
62
average
1.84
0.96
1.2
Erms
1.90
1.41
1.57
Percent Accuracy*
7
74%
73%
86%
80%
83%
78%
* The accuracy of predicting the 7 sequence pattern is calculated from the average and Erms values by dividing by 7 and subtracting from 100%.
With this analysis we can see that none of the 7 sequence patterns are a good match to the episodes ... even though the previous analysis looked good (but did not properly account for |departure| from the pattern) ... this essentially predicts 1 to 2 errors of matching an episode to a specific 7 sequence pattern ...
Curiously I do not find this a compelling "pattern" in this regard. Nor do I see any value in lumping the three 7 sequence patterns into an uber pattern.
. Good to know you are not one of those people that cannot admit to errors.
so NOW we have (updating the pattern in Message 166:
™ Version 4.2
(1)