On Jul 4, 4:44 pm, gimpe...@hotmail.com wrote: > Ray: > > In case you're occasionally checking sci.stat.math while you're out-of- > pocket, I'm posting the following results now rather than later. > > The paired "Y's" in the following table indicate that: > > a) we ARE justified in using doubly-dichotomized residuals from our > two linear regressions (ln(c/u) on ln(c/L) and ln(c/e) on ln(c/L)) as > predictors of structural alignability in our intermediate logistic > regression; > > b) but - we have to be very careful about where we get the residuals, > i.e from runs of the two regressions on WHAT dicodon sets with WHAT > restrictions on the value of "u" (representation level). > > A detailed legend follows the table - its complex, but will make the > table comprehensible. > > Results of Significance Testing > with Custom Heteroscedastic t-test > After Intra-Family Bonferroni Correction > > Fold a1 > > F1 F2 F3 > Test u e u e u e > > 1a) uL for S vs uL for S_R * > 1b) uL for C vs uL for C_R * Y Y > 2a) uH for S vs uH for S_R * Y Y * > 2b) uH for C vs uH for C_R > 3a) uA for S vs uA for S_R * Y Y > 3b) uA for C vs UA for C_R * Y Y > > 4a) uL for S vs uL for C d > 4b) uL for S_R vs uL for C_R r r r > 5a) uH for S vs uH for C Y Y * > 5b) uH for S_R vs uH for C_R > 6a) uA for S vs uA for C * d > 6b) uA for S_R vs uA for C_R r r r > > 7a) uL vs uH for S * Y Y > 7b) uL vs uH for S_R > 8a) uL vs uH for C d d d d d > 8b) uL vs uH for C_R r r r r r r > > Here is the legend needed to understand the table: > > I) Identity of S, C, S_R, and C_R dicodon sets used in test suite > Familes F1-3: > > |-----TestSuite Family----| > Dicodon Set F1 F2 F3 > S S63 S119 S60 > C C711 C1058 C493 > S_R S63R S63R S63R > C_R C673R C673R C673R > > (So, for example, in Families F1-3, test 1a was performed using > dicodon sets (S63,S63R), (S119,S63R),and (S60,S63R) respectively.) > > II) Meaning of "u" and "e": > > "u" : ln(c/u) on ln(c/L) regression (representation-level regression) > "e" : ln(c/e) on ln(c/L) regression (enthalpic-level regression) > > III) Meaning of "uL", "uH", and "uA": > > uL: regression was run on observations with u < 0 <= 1 > uH: regression was run on observations with u > 1 > uA: regression was run on observations with u > 0 > > IV) Meaning of cell-entries "*", "r", "d", and "Y" > > A) All of "*", "r", "d", and "Y" indicate that your custom > heteroscedastic t-test was significant for BOTH > slope AND intercept AFTER intra-family Bonferroni correction. > > (Recall that this t-test considers the results of running a given > regression on our 12 agree-upon length intervals.) > > B) "r" indicates the result was obtained solely from RANDOM data; > > C) "d" indicates that although the result was obtained solely from NON- > random data, its significnce must be disregarded because the result > of the corresponding test on RANDOM data was also signficant; > > D) "*" indicates that: > > 1) the result was obtained solely from NON-random data (or non- > random vs random data); > > 2) the corresponding test on solely random data was NOT > significant; > > 3) BUT the result was obtained only for the "u" regression, or the > "e" regression, not both; > > E) "Y" indicates that: > > 1) the result was obtained solely from NON-random data (or non- > random vs random data); > > 2) the corresponding test on random data was NOT significant; > > 3) AND the result was obtained for BOTH the "u" regression and the > "e" regression.