It depends on what you mean by "interaction" but a usual method is to combine the six questions into one; there are different methods of doing this depending on the nature of the questions you are combining, but factor analysis is often useful.
I have done a study on whether personality and demographics predict interaction on facebook brand pages.
I have used a Big five personality scale and the demographics include, sex, age, marital status, employment status and education level.
Interaction was measured using 6 questions, such as 'how many brand pages have you liked in total', ' how many times per week do you comment on a brand pages post' etc.
So far I have done a pearsons correlation so I can see where the positive and negative correlations are.
Which statistical test would be best to use? I am using SPSS. I was thinking about doing a multiple regression but the only way I could do it is by doing 6 different ones for each D.V (interaction).
Thank you for any help
It depends on what you mean by "interaction" but a usual method is to combine the six questions into one; there are different methods of doing this depending on the nature of the questions you are combining, but factor analysis is often useful.
Hi thanks for replying!
Interaction on facebook brand pages was measured using 6 different questions.
1)how many total brand pages have you liked? 0-5, 6-11, 12-20, 21-30, over 30
2)how much time per hour do you spend using facebook brand pages? under 1 hour, 1-2 hours, 2-4 hours, 4-6 hours, 6+ hours
3)how many times per week do you comment on a brand pages post? (answers from 0 to 6+)
4)how many times per week do you share a brand pages post (answers from 0 to 6+)
5)how many times per week do you like a brand pages post (answers from 0 to 6+)
6)how many times per week do you post on a brand pages wall (answers from 0 to 6+)
The data is all ordinal. Would factor analysis be suitable?
Well, these are questions which should be asked
during the design stage of the study, and not only
after data collection has already been done.
Anyway, you cannot perform Pearson correlations
with these variables, since they are clearly ordinal.
You also cannot use linear regression, for the same reason.
Whether it is possible to aggregate the 6 variables,
I don't know. Factor analysis can't be used, AFAICS.
At least you could think about 6 multiple ordinal
regressions.
With kind regards
K.
It has been a while, but I believe factor analysis can use spearman's or polychoric correlations either of which might make more sense in this case. When I was in school polychoric got more attention for likert scale data and that is how I ran my factor analysis last time I used this data I believe.
You can always run ordered logistic regression although it is not clear to me exactly what you are doing. Are you trying to predict an ordered DV, collapse categories...
"Very few theories have been abandoned because they were found to be invalid on the basis of empirical evidence...." Spanos, 1995
Factor analysis can be done on ordinal data, but there are some additional steps. Joreskog & Sorbom have an excellent article on this: http://stat-athens.aueb.gr/~moustaki...les/paper7.pdf
Tweet |