Advice requested about mixed ANOVA

#1
Need your help with an Anova analysis

Hi

I'm studying the effect of different music styles on performance tasks.
I'm considering a mixed ANOVA analysis, but I'm not very confident about my stats skills and I'd be very happy to receive good advices from the experts here.

Subjects are split (randomly) in 3 groups, based on the music style of a song (that they listen, in the experimental condition). So this condition (3 modalities : MS1, MS2, MS3) would be my between-subjects factor.

Performance is measured before and after listening to the track. Its the same performance task for all subjects. So Time (2 measures : T0, T1) would be my within-subjects factor.

Now this is what I'd like to test :

- Is there a performance improvement between T0 & T1, depending on the music style.

but more importantly :

- Is there a significant difference in the improvement between T0 & T1 depending on the music style.

In other words, is there less improvement between T0 & T1 for the performance in MS1 condition compared to the other conditions for example.

For this last one, a mixed ANOVA is especially required I guess, but I'm not sure about the way to deal with the less/more improvement aspect. Should I use the contrast or post-hoc for example ?

Thank you very much for any input.
 
Last edited:
#2
Re: Need your help with an Anova analysis

If you are interested mainly in the question of whether or not the three groups show different improvements, you would be looking for a significant Time/Modality interaction, Interactions with more than two groups are hard to interpret and explain.
An alternative, and mathematically equivalent, is to find the before/after difference for each subject and do a one way anova on the differences. Easier to do, interpret and explain.
 
#3
Re: Need your help with an Anova analysis

If you are interested mainly in the question of whether or not the three groups show different improvements, you would be looking for a significant Time/Modality interaction, Interactions with more than two groups are hard to interpret and explain.
An alternative, and mathematically equivalent, is to find the before/after difference for each subject and do a one way anova on the differences. Easier to do, interpret and explain.
Hello Katxt.

If I create a before / after data for each subject (I guess that it would be " value of performance at T1" minus "value performance at T0") then I could indeed do a one way anova with that. If I understand correctly, then it should be a "between-subject" anova, right ?
Its smart, indeed. Thank you very much!
 
#4
Re: Need your help with an Anova analysis

That's right. A one way anova is always between subjects. Many stats programs also give you a p value for the constant (the average value). This would answer your first question about whether there was a time difference overall.