Re: st: Re: Re: Repeated Measures ANOVA: contrasts

[Vincent Koppelmans <vkoppel@umich.edu>]

Dear Dr. Coveney,

Thank you very much for the syntax and the follow up email.

I have a few questions:

1) What is the difference between the contrasts: 
	1.5 1.5 -1 -1 -1 0 0
	0.5 0.5 -1/3 -1/3 -1/3 0 0

I thought the would be the same as long as the weights are proportional?



2) You suggest to use a paired t-test on contrast variables (i.e., (time point 1 + 2) - (time point 3 + 4 + 5)).
If I understand correctly, you are saying that manova/contrast designs are not valid (because of lack of power?)?

The 7 time points I have can be grouped as follows:
Time point 1 and 2: pre treatment
Time point 3, 4, and 5: treatment
Time point 6 and 7: post treatment

Hence, I do not expect a difference between TP 1 and 2, but I do expect some cumulative effect from time point 3 to 5.

This is why I wanted to use contrasts instead of calculating averages per condition (i.e., pre, during, and post treatment).

What would be a good and valid solution here?
 
Thank you for your time!

best,

Vincent



Op 2 feb. 2014, om 04:36 heeft Joseph Coveney <stajc2@gmail.com> het volgende geschreven:

> David C Airey wrote:
> 
> I didn't read far enough in the help file! Page 362 and beyond
> for -contrast- has examples with evaluated fractions like
> `=1/3' in the linear combination.
> 
> --------------------------------------------------------------------------------
> 
> I didn't read that far either.  Anyway, if there is substantial autocorrelation,
> and if the difference between the average of the first two observations and the
> following three is of primary scientific  interest, then the original poster
> would perhaps be better off just computing the averages for each volunteer and
> performing a paired t-test.  -manova- and -manovatest- (or -contrast-) are out,
> with eight volunteers and seven intervals.
> 
> Joseph Coveney
> 
> . set seed `=date("2014-02-02", "YMD")'
> 
> . 
> . program define simem, rclass
>  1.     version 12.1
>  2.     syntax
>  3. 
> .     drop _all
>  4.         set obs 8
>  5.     generate byte pid = _n
>  6.     generate double fmt1 = rnormal()
>  7.     forvalues i = 2/7 {
>  8.         generate double fmt`i' = 0.5 * fmt`=`i'-1' + ///
>>            sqrt(0.75) * rnormal()
>  9.     }
> 10.     tempvar a b
> 11.     generate double `a' = (fmt1 + fmt2) / 2
> 12.     generate double `b' = (fmt3 + fmt4 + fmt5) / 3
> 13.     ttest `a' = `b'
> 14.     tempname p
> 15.     scalar define `p' = r(p)
> 16.     drop `a' `b'
> 17.     quietly reshape long fmt, i(pid) j(time)
> 18.     xtreg fmt i.time, i(pid) fe
> 19.     test (1.time + 2.time) / 2 = (3.time + 4.time + 5.time) / 3
> 20.     return scalar l = r(p)
> 21.     return scalar t = `p'
> 22. end
> 
> . 
> . simulate t = r(t) l = r(l), reps(10000) nodots: simem
> 
>      command:  simem
>            t:  r(t)
>            l:  r(l)
> 
> 
> . 
> . foreach var of varlist l t {
>  2.         generate byte pos_`var' = `var' < 0.05
>  3. }
> 
> . 
> . summarize pos_*
> 
>    Variable |       Obs        Mean    Std. Dev.       Min        Max
> -------------+--------------------------------------------------------
>       pos_l |     10000        .105    .3065687          0          1
>       pos_t |     10000       .0511    .2202127          0          1
> 
> . 
> . exit
> 
> end of do-file
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/