Which Variables to use for Repeated Measures ANOVA? (in R)

#1
Hi all,

I am currently doing research on the relationship between music and emotions, as part of my PhD.

I have conducted a musical stimuli validation study, where 28 short musical tracks were composed, each composed with the intention of conveying a specific emotion to its listeners. To validate that each musical track is actually conveying the intended emotion, I carried out an online survey where participants (N=96) were asked to listen to each track and rate on Likert scales how much of 9 emotions they thought the music was expressing.

My supervisor instructed that I have to analyse the data using repeated measures ANOVA, and possibly also go into MANOVA.

My variables are the following:
Independent variable: music tracks (N=28)
Dependent variable: emotion rating

My issue is that the dependent variable was rated on 9 different scales. So for each track, 9 ratings were given.
For one track, participants had to rate how much of sadness, happiness, longing etc. (9 emotion scales in total) they thought the music is expressing. Furthermore, each track has a 'Target Emotion' assigned to it (which is another variable), and the aim of this experiment is to discover if the Target Emotion for each track was selected by the participants.

My question is if I am correct in using the music tracks as the independent variable and what my dependent variable should be as there are multiple ratings for each independent variable?

Any help would be appreciated, as I am quite stumped.

Regards,

Annaliese
 

Karabiner

TS Contributor
#2
rate on Likert scales how much of 9 emotions they thought the music was expressing.
Not 9 Likert scales, but 9 single Likert-type items, I suppose? Or maybe just some rating items,
not even Likert-type items? How does the answering format of these items look like, exactely?
These consideration are sometimes important regarding the level of measurement of variables
(ordinal vs. interval or quasi-interval).

My supervisor instructed that I have to analyse the data using repeated measures ANOVA, and possibly also go into MANOVA.
I am not sure what should be achieved by this. You say you're doing a validation study,
then why hypothesis testing?

My question is if I am correct in using the music tracks as the independent variable and what my dependent variable should be as there are multiple ratings for each independent variable?.
See above - what exactely do you want to find out, what is the aim of your analysis?

With kind regards

Karabiner
 
#3
Dear Karabiner,
Thank you for your reply.

How does the answering format of these items look like, exactely?
Please find attached what a question and answering format looks like.

what is the aim of your analysis?
The aim of the study is to see whether the composer's intended emotion in each track was successfully conveyed to the listeners. So for each track, there is a Target Emotion, and what I want to do is analyse the listeners' ratings and identify if the mean ratings of the 'correct' target emotion for each track were significantly different from the mean ratings of the other emotions.

Thank you so much for your help.

Best wishes,

Annaliese
 

Attachments

#4
This is going to be difficult to interpret I think. My concern would be multicollinearity with the data because of all the emotions measured. However, it's not impossible.

What type of rating do you have for the 'correct emotion'? I think you need a comparison scale to compare your scores to if you are going to do this, because the composer saying 'This is a sad song' isn't a variable but it is an intent. However if participants feel that the song is a happy song then you don't have a metric to compare against. You may be able to do something along the lines of 'The composer said this was a sad song, and X people rated this song as a sad song. Compared to everyone else who rated the song as (describe), the sad song people also described the song as (describe)." Not entirely sure if this is what you are looking for though.

I can't speak for sure, but I don't know if repeated measures ANOVA is the best choice for analysis. Your participants completed the same measure but for different stimuli (independent variable is different). MANOVA may be the way to go because you are comparing the effect of different musical pieces on the 9 different emotions.

I would use the musical pieces as groupings and analyze the differences between the musical pieces in terms of emotional response.

X = Music
Y = Emotion

Pretty sure this is MANOVA, but my experience with it is very limited so I can't provide much insight there.

Hope this helps.
 
#5
Thank you for your reply gdaem, it is very helpful.

I would use the musical pieces as groupings and analyze the differences between the musical pieces in terms of emotional response.
I think I managed to work around it kind of; for each track X, I calculate the mean ratings of all the 9 emotions. If the 'target emotion' (Y?) rating is significantly higher than the other emotions, I can say that the intended emotion has been conveyed to the listeners.
What type of rating do you have for the 'correct emotion'?
There isn't a specific rating for the 'correct emotion', as in, there isn't a specific value that I can compare the mean ratings of the participants to. I think as long as the highest emotion rating is that of the target emotion, I think I can say that that is what I was looking for?

I will look into MANOVA.

Thank you once again for your help!
 
#6
Be careful with that last one. If you state that your highest value is the one you were looking for, you run the risk of readers or committee members viewing it as HARKing (Hypothesizing After Results Known). Provided it is worded correctly and your analyses are set up and interpreted in a way to avoid this, I think you've got a good hypothesis!
Glad to help! :)