How to report result in Anova repeated measure

[Tommaso Salvitti]

Good morning to everybody,
how to write the result of ANOVA REPEATED MEASURES?



I have attached the example in the "Example" txt file.

Thank you to everybody
.

[Tommaso Salvitti]

dataex


*ACmean
. anova Acmean Outcome_new_num##Timepoint_new ident, repeated (Timepoint_new) bseunit(ident) bse(Outcome_ne
> w_num)

Number of obs = 334 R-squared = 0.3027
Root MSE = .074427 Adj R-squared = -0.0413

Source | Partial SS df MS F Prob > F
------------------------+----------------------------------------------------
Model | .536255969 110 .004875054 0.88 0.7733
|
Outcome_n~m | .000310956 1 .000310956 0.06 0.8129
Timepoint~w | .03772893 4 .009432232 1.70 0.1503
Outcome_n~m#Timepoint~w | .069893848 4 .017473462 3.15 0.0151
ident | .418709517 101 .004145639 0.75 0.9507
|
Residual | 1.23529175 223 .005539425
------------------------+----------------------------------------------------
Total | 1.77154772 333 .005319963

[Tommaso Salvitti]

Nick Cox can you help me please???

[Nick Cox]

Sorry, no. Repeated measures ANOVA I have never done.

[Tommaso Salvitti]

I don't believe that nobody can help me how to write the result..
However thanks to everybody

[Joseph Coveney]

I can't help you write the result, but I have a few recommendations:

1. For a repeated-measures ANOVA, you'd be wanting a so-called balanced dataset in order for the test (F) statistics to be distributed as they're supposed to be. With 102 subjects, two levels of the grouping (between-subjects) factor and five of the repeated (within-subjects) factor, a balanced dataset would have 102 × 5 = 510 observations. You have only 334, which implies that you have a very unbalanced dataset. You really shouldn't be using -anova- at all. Look into an equivalent syntax for repeated-measures ANOVA for the iterative maximum likelihood estimation command, -mixed-.

2. Why are you using what looks like a dependent variable ("Outcome_new_num") as the between-subjects factor, that is, as an independent variable?

3. You're using what appears to be a summary statistic ("Acmean") as your dependent variable. This is liable to give you a misleadingly optimistic estimate of your uncertainty. I recommend modeling the individual responses as your outcome (dependent) variable in order to get a better picture of where your errors lie.

Last, I'm not sure about others on the List, but I didn't quite understand what it is that you're looking for exactly when you ask for help "to write the result". Are you looking for assistance in interpretation? That would require substance matter knowledge, wouldn't it? "Acmean", "Outcome_new_num", "Timepoint_new" and "ident" don't give much clue to just what that is. Otherwise, for help in interpretation, about all that I can suggest is to take a look at -margins- and -marginsplot- for profile plots.