Topic:
The same four proportional weighting factors work for each 00/01/10/11 when 0.25 is subtracted from each !!!
Replies:
506
Last Post:
Nov 20, 2012 9:21 PM
On Jun 19, 5:42 am, djh <halitsk...@att.net> wrote: > All a1 data is now complete for: > > 3 dicodon sets (S63, C711, S63R) > 12 length intervals (indexed 1-12) > 2 types of c (simplified and averaged) > 2 types of regressions (ln(c/u) on ln(c/L) and ln(c/e) on ln(c/L) > > So there are 144 different regressions in all, and I will be sending > you offline the intercept, s_e(intercept), slope, and s_e(slope for > all of them in one csv file with the following header line > > Fold (for future cross-fold comparisons > Dicodon Set > Length interval index > Type of "c" > Regression Type > intercept > s_e(intercept) > slope > s_e(slope) > > Please note that the intercept, slope, and their s_e's will be as > reported by Excel - I do not yet have access to Minitab or the other > "cheap but good" package I mentioned to you. > > I will also be sending you one detail file with the same "ID" > information in the header (plus unique observation ID) > > Fold (for future cross-fold comparisons > Dicodon Set > Length interval index > Type of "c" > Regression Type > Observation ID > > and the following detail information > > Actual L > Simplified c > Averaged c > "e" > "u" > > Please note that "observation ID" will not be unique across the whole > file (since the same chain segment can be selected for more than one > of the 144 categories), but it will be unique within > > Fold (for future cross-fold comparisons > Dicodon Set > Length interval index > Type of "c" > Regression Type > > However you choose to do your analyses, what we first need to know is: > > a) where there are significant differences between the behavior of the > two correlations on S63 data vs S63R data > b) where there are significant differences between the behavior of the > two correlations on S63 data vs C711 data > c) whether all the differences noted in (a,b), taken together, are > strong enough to permit us to use the results of the two correlations > on the S63 data as drivers for the logistic regressions (as we have > been doing) > > We also need to know whether: > > d) averaged c tracks closer to simplified c in more S63 cases than > C711 cases or S63R cases > > particularly since this may wind up to be a main distinguishing > characteristic of S63 vs C711 and S63 R (if you think it's legitimate > to use such a difference as a distinguishing characteristic.) > > In general, please do not expect the majority of "parallel" > comparisons to show NO CI overlap. There are cases of: > > e) no CI overlap > f) very small CI overlap > g) greater CI overlap but still significant sections of non-overlap > h) cases where one CI is entirely contained by another > "symmetrically" (i.e. one is centered within the other) > i) cases where one CI is entirely contained by another, but one is > either at the low-end of high-end of the other. > > So I am hoping that your "more sophisticated" tests will tell us that > we can use the results of the two correlations on S63 data, i.e. your > tests that do not depend simply on the consistent non-existence of > overlaps. > > The detail and summary data files should go out sometime tonight. If > you happen to see this post sometime today, and want me to change the > way I'm intending to send you data, please let me know at your > earliest convenience.
Fold a1 Average Average Intercept Slope Pairwise Contrasts
e s S63 2.87534 1.38088 All avg intercepts e s C711 2.54335 1.18938 and all avg slopes e s S63R 2.04731 0.911252 differ significantly.
e a S63 3.02306 1.47368 Same as above. e a C711 2.53308 1.18089 e a S63R 2.29048 1.06117
u s S63 3.75309 1.00594 No slope diffs are sig. u s C711 4.04267 0.962134 Only S63-C711 int diff u s S63R 4.02223 0.971684 is sig.
u a S63 3.68991 0.963629 No slope diffs are sig. u a C711 4.06502 0.977247 Both S63-C711 & S63-S63R u a S63R 4.23071 1.09593 int diffs are sig.
There were no significant pairwise differences between data sets in terms of the patterns of their intercepts or slopes across the length intervals.
