Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    You might try
    Code:
    mixed c.age i.sex i.trt##i.week || pid: , reml dfmethod(satterthwaite)
    and see whether that gets you where you want to go.

    Comment


    • #17
      It might help you to run the following do-file to see what's going on with -mixed- and Stata's -anova- for repeated-measures ANCOVA. (The analyses start at the "Begin here" comment; the first part is just to create a balanced dataset for illustration.)

      Compare results from these approaches using your dataset with what your former colleague got with SPSS.
      Code:
      version 14.2
      
      clear *
      set more off
      set seed 1384471
      quietly set obs 52
      
      generate byte pid = _n
      generate double u = rnormal()
      
      generate byte sex = mod(_n, 2)
      generate byte age = runiformint(18, 65)
      
      bysort sex (pid): generate byte trt = mod(_n, 2)
      
      quietly expand 2
      bysort pid: generate byte week = _n - 1
      generate double y = u + rnormal()
      
      *
      * Begin here
      *
      
      /* Repeated-measures ANCOVA using -mixed-
         (Suggested in post immediately above in this thread) */
      mixed y c.age i.sex i.trt##i.week || pid: , reml dfmethod(satterthwaite) nolrtest nolog
      
      // Tests of ANOVA main effects (i.e., ANOVA parameterization of main-effects contrasts)
      contrast trt, small
      contrast week, small
      contrast sex, small
      
      /* Repeated-measures ANCOVA using -anova-
         1. As per previous posts in this thread, ignore the test of the age covariate)
         2. Also, -repeated(week)- is not needed here because there are only two levels of the week factor */
      anova y c.age sex trt / sex#trt|pid week trt#week
      
      // To test age covariate (if you really care):
      
      * Either using -xtreg , be-
      xtreg y c.age i.sex i.trt##i.week, i(pid) be
      
      * (Side note:  see also the following for repeated-measures factors)
      xtreg y c.age i.sex i.trt##i.week, i(pid) fe
      contrast week // (Again, ANOVA parameterization of main-effects contrast)
      
      * or using -manova- followed by -manovatest-
      quietly reshape wide y, i(pid) j(week)
      quietly manova y0 y1 = c.age i.sex i.trt
      
      matrix define M = J(1, 2, 1)
      
      manovatest , showorder
      matrix define A = (1, 0, 0, 0, 0, 0)
      
      manovatest , test(A) ytransform(M)
      
      exit

      Comment

      Working...
      X