To give more background, what I have is a personality trait measure (giving a score for each participant) and the answers each participant gave on 20 trials, for which their answers were from a nominal list (e.g. reason 1, reason 2, ... reason 7).
The experiment was fully within-participants. Another item is that out of the 20 trials, 10 were of a certain type (e.g. blue) and 10 were a different one (e.g. red).
How to I analyse if individuals that are higher or lower on that personality trait are more/less likely to select specific reasons for the trials, while also considering type of trial?
(I was thinking a multinomial regression, but I have no idea how to split the trials in SPSS).
Thanks in advance!
Dear Mircea,
One place you could start is binary logistic regression (as it is complicated enough in itself!), maybe with your two most popular reasons for the first trial. You could analyse the 20 trials separately in this way. You could then scale up your analysis to multinomial logistic regression (which is essentially just an extension of binary) if you need to, but watch out for low frequencies in responses..
Hi Peter,
Thank you for the reply. I will consider that approach.
I was also considering tallying the responses of each participant over the 20 trials (or 10 and 10, based on the additional factor), and then [somehow] regressing the frequency of each of the 7 reasons with the personality score.
Any idea if this sounds like a viable statistical approach?
So lets make the situation clearer.
Do you have 20 different trait scores from a person?
I suppose you used a likert type scale with 7 choices.
There is a trial factor which has two levels: blue and red.
And there is a time factor which has 20 levels.
Is that correct?
There were 20 trials in total, half blue, half red (within-participants).
On each trial participants could choose 1 of 7 reasons (nominal).
After the experiment was over, the also completed a personality questionnaire (continuous).
So now I want to analyse any relationship between that personality score and choice in reason, while also accounting for if the trial was a red or blue trial.
So 20 ,say, questions (half red, half blue) are asked to a subject. And he or she gives 20 nominal responses. And after that you measure the trait, So one score for trait. At total you have 21 scores for each subject. Is that correct?
You want to analyze relationship between personality score and nominal responses while accounting for the trial is blue or red.
How many participants do you have?
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS