[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

From |
tmmanini <tmmanini@mailbox.syr.edu> |

To |
Statalist <statalist@hsphsun2.harvard.edu> |

Subject |
RE: st: Re: ancova for repeated designs |

Date |
Mon, 16 Aug 2004 00:25:12 -0400 |

You have all been very helpful, thank you. You are right that I have only 6 levels of convariate (a possible problem), but I took your advice on several fronts and I'm still not fully comprehending the solution. Here's what I did: (I used my data with 32 subjects, which is included at the end). First I ran the model positioning g after the id|g random error term, and the specifying if t>1. I got a sig. interaction, but according a recent addition to the listserv, I learned that this interaction may not be as important as I once thought. Therefore, I dropped the g*t term from the model. anova y g / id|g x g*x t g*t if t>1, rep(t) cont(x) Number of obs = 64 R-squared = 0.8051 Root MSE = 1.25373 Adj R-squared = 0.5767 Source | Partial SS df MS F Prob > F -----------+---------------------------------------------------- Model | 188.354167 34 5.53982843 3.52 0.0004 | g | .321850082 2 .160925041 0.07 0.9337 id|g | 60.8061356 26 2.33869752 -----------+---------------------------------------------------- x | 5.33333333 1 5.33333333 3.39 0.0757 g*x | 10.6666667 2 5.33333333 3.39 0.0474 t | 22.0119048 1 22.0119048 14.00 0.0008 g*t | 2.16666667 2 1.08333333 0.69 0.5100 | Residual | 45.5833333 29 1.57183908 -----------+---------------------------------------------------- Total | 233.9375 63 3.71329365 g*x dropped from the model anova y g / id|g x t g*t if t>1, rep(t) cont(x) Number of obs = 64 R-squared = 0.8051 Root MSE = 1.25373 Adj R-squared = 0.5767 Source | Partial SS df MS F Prob > F -----------+---------------------------------------------------- Model | 188.354167 34 5.53982843 3.52 0.0004 | g | 12.0364118 2 6.01820591 2.53 0.0977 id|g | 66.5887517 28 2.3781697 -----------+---------------------------------------------------- x | 2.7111e-28 1 2.7111e-28 0.00 1.0000 t | 22.0119048 1 22.0119048 14.00 0.0008 g*t | 2.16666667 2 1.08333333 0.69 0.5100 | Residual | 45.5833333 29 1.57183908 -----------+---------------------------------------------------- Total | 233.9375 63 3.71329365 These results seemed weird, based on the previous F value for x being much higher. So I dropped t>1 from the model anova y g / id|g x t g*t, rep(t) cont(x) Number of obs = 96 R-squared = 0.7789 Root MSE = 1.18014 Adj R-squared = 0.6379 Source | Partial SS df MS F Prob > F -----------+---------------------------------------------------- Model | 284.628472 37 7.69266141 5.52 0.0000 | g | 8.02427455 2 4.01213727 2.53 0.0977 id|g | 44.3925011 28 1.58544647 -----------+---------------------------------------------------- x | .166666667 1 .166666667 0.12 0.7306 t | 34.015873 2 17.0079365 12.21 0.0000 g*t | 5.63888889 4 1.40972222 1.01 0.4087 | Residual | 80.7777778 58 1.39272031 -----------+---------------------------------------------------- Total | 365.40625 95 3.84638158 I'm not sure which model it correct? Based on recent addition by Joseph Coveney, the last model (without t>1) would be correct. Here is the data, sorry it is long, there are 32 subjects, 3 levels of g, 3 levels of t and 1 level of x (remeber x is the first level of t (time)) I'm trying to covary for the pre-test level (time==1). One more thing, I successfully implemented the adjust command by included id in the "by" statement. However, I only received adjustments for those subjects I specify (ie. id<=4 gives me subjects 1 through 3), which makes sense. However, I would like to report the adjusted mean for each group over each time period. I guess I can request all id's be shown on the output by using "adjust x, by(g t id)" and then taking the mean of the id's for each group, but that seems cumbersome. Is there a better way? By the way thank you again for all your help. id g t y x 1 1 1 1 1 1 1 2 1 1 1 1 3 1 1 2 2 1 1 1 2 2 2 1 1 2 2 3 1 1 3 3 1 5 5 3 3 2 5 5 3 3 3 1 5 4 2 1 5 5 4 2 2 4 5 4 2 3 1 5 5 3 1 6 6 5 3 3 6 6 5 3 2 6 6 6 1 1 3 3 6 1 2 3 3 6 1 3 3 3 7 1 1 6 6 7 1 2 6 6 7 1 3 6 6 8 1 1 1 1 8 1 2 1 1 8 1 3 1 1 9 2 1 5 5 9 2 2 6 5 9 2 3 1 5 10 1 1 1 1 10 1 2 1 1 10 1 3 1 1 11 1 1 4 4 11 1 2 4 4 11 1 3 1 4 12 2 1 2 2 12 2 2 2 2 12 2 3 2 2 13 2 1 5 5 13 2 2 1 5 13 2 3 1 5 14 1 1 1 1 14 1 2 1 1 14 1 3 1 1 15 1 1 5 5 15 1 2 5 5 15 1 3 1 5 16 3 1 1 1 16 3 2 1 1 16 3 3 1 1 17 1 1 6 6 17 1 2 6 6 17 1 3 6 6 18 2 1 2 2 18 2 2 2 2 18 2 3 1 2 19 3 1 4 4 19 3 2 4 4 19 3 3 2 4 20 2 1 1 1 20 2 2 1 1 20 2 3 1 1 21 1 1 6 6 21 1 2 6 6 21 1 3 1 6 22 3 1 6 6 22 3 2 6 6 22 3 3 1 6 23 2 1 1 1 23 2 2 1 1 23 2 3 1 1 24 2 1 5 5 24 2 2 5 5 24 2 3 4 5 25 2 1 5 5 25 2 2 1 5 25 2 3 1 5 26 2 1 2 2 26 2 2 1 2 26 2 3 1 2 27 3 1 2 2 27 3 2 2 2 27 3 3 1 2 28 2 1 1 1 28 2 2 1 1 28 2 3 1 1 29 1 1 1 1 29 1 2 1 1 29 1 3 1 1 30 3 1 3 3 30 3 2 3 3 30 3 3 3 3 31 1 1 2 2 31 1 2 5 2 31 1 3 5 2 32 3 1 4 4 32 3 2 4 4 32 3 3 2 4 >===== Original Message From Joseph Coveney <jcoveney@bigplanet.com> ===== >An example of a time-invariant repeated measures ANCOVA is shown below. It >is from B. J. Winer, D. R. Brown and K. M. Michels, _Statistical Principles >in Experimental Design_ Third Edition (New York: McGraw-Hill, 1991), >pp.828-832. The do-file reproduces the results in the text within rounding >error (and after correcting a typographical error in the text in >Table 10.34). > >As I mentioned yesterday in a manner that was incomprehensibly articulated >("there is no variation of x within id, so there won't be any within the >id|g error term, either, and it should be put to the right of the id|g >random error term"?), the continuous covariate shouldn't share the >within-subjects error term with the between-subjects factor, and should be >moved to the right of the id|g term. Winer's example does not include a >term for the covariate-by-between-groups-factor interaction. For most >purposes, between-group homogeneity of the slope of the continuous covariate >is assumed, since its violation probably couldn't be powerfully detected by >the statistical significance of the interaction term for most datasets not >specifically powered to examine the interaction. > >Joseph Coveney > >clear >set more off >input byte subj byte a byte x1 byte y1 byte x2 byte y2 >1 1 3 10 3 8 // Note typographical error in text's Table 10.34 >2 1 5 15 5 12 >3 1 8 20 8 14 >4 1 2 12 2 6 >5 2 1 15 1 10 >6 2 8 25 8 20 >7 2 10 20 10 15 >8 2 2 15 2 10 >end >reshape long x y, i(subj) j(b) >/* Unadjusted repeated-measures ANOVA--Part (i) of Table 10.35 > on Page 830 */ >anova y a / a|subj b a*b, repeated(b) >/* Repeated-measures ANCOVA--Part (ii) of Table 10.35 */ >anova y a / a|subj x b a*b, repeated(b) continuous(x) >adjust x if subj==1 | subj==5, by(a b subj) >exit > > > > > >* >* For searches and help try: >* http://www.stata.com/support/faqs/res/findit.html >* http://www.stata.com/support/statalist/faq >* http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

- Prev by Date:
**st: date translation** - Next by Date:
**st: RE: missingness in a large, complex sampling deign** - Previous by thread:
**st: Re: ancova for repeated designs** - Next by thread:
**st: Re: ancova for repeated designs** - Index(es):

© Copyright 1996–2014 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |