How many groups of people are you considering?
We had individuals fill out a survey where they assign importance to five categories. So for example: if given $1000 to distribute among five different categories, where more money entails more importance, how would you distribute your $1000 among the following five categories?
Then we just provide five different categories for them to "rank," so that we have group means for how much money people give to category 1, category 2, etc.
My question is: what sort of ANOVA (I assume ANOVA) are we to use to determine significant differences among these categories with continuous data? Basically asking: "did our group significantly prefer one category over others?"
Sorry for the dumb question - haven't used ANOVA in years.
How many groups of people are you considering?
hi,
I think that a simple one way ANOVA should work - the twist is of course that the five categories are not independent , they always sum to 1000. I have not seen the requirement that all groups need to be independent in ANOVA though , but given the way ANOVA is used this could be something everybody takes for granted.
Thinking about the math, it seems to menthat the fact that the values sum to a constant should not be a problem, but I may be wrong.
regards
rogojel
It's just one group, aiming for about 1,000 people.
One of the parts I've been struggling with is determining IV and DV when it comes time to running this in SPSS. Sounds simple enough, but it's really had me thinking.
Basically we have five categories which are being given a certain weight (or in this case, being given fake money). What we essentially want to do is figure out if any of the categories are significantly favored compared to the others. So literally it's like:
Category 1 = "some mean amount"
Category 2 = "some mean amount"
Category 3 = "some mean amount"
Category 4 = "some mean amount"
Category 5 = "some mean amount"
And we are comparing the mean money given to each category. But how do you identify IV and DV in that? I wasn't sure if I would need a nonparametric approach, which I'm not super familiar with.
Well,I am not familiar with SPSS but this should be one factor with 5 levels. It could be represented by two columns , in one you have the level IDs ( like a number of 1s for level 1 2s for level 2 ) and a next column where each entry is a ranking given for the corresponding level.
level ranking
1 100
1 956
-----------------
2 220
ezc
regards
rogojel
What would the DV be, though? I understand what you're saying conceptually, but I still feel a bit lost on how to analyze for significant differences among categories.
As global test you could use a repeated measures ANOVA on the
amounts of money (= DV), with "category" as within-subjects factor
(= IV). Or, if you are just interested in the ranking, not in the
mean differences between categories, then just use a Friedman
test.
Wih kind regards
K.
Tweet |