A simple analysis would just compare the mean rank between the ten options. A more sophisticated analysis might use a multilevel logistic regression to model the rank as a function of gender and reasons for moving.

However, there might be other ways to approach such data - for example through some type of clustering approach - depending on your research question.

You have ipsative measures (the sum must = 1+2+3+4+5+6+7+8+9+10). Greer & Dunlap have argued one can run ANOVAs on such measures. Just run a repeated measures ANOVA : Gender (2 level if this between factor) x Reasons (10 levels of this within/repeated measures factor)

Citation:

Greer, T., & Dunlap, W. P. (1997). Analysis of variance with ipsative measures. Psychological Methods, 2(2), 200-207 .

If I understand your question correctly, the outcome, for each participant, should be a permutation of the numbers from 1 to 10, since they are ranking their 10 reasons for moving abroad (1- from most preferred reason to least ). Keep in mind that you present all of them the same LIST of reasons and they put the first 10 digits according to their preferences. This type of outcome is sometimes called preference data or rank data. I think the best way to analyze it, is first put the data in long format (you will have 10 lines of data for each participant) and, as Thom suggests, use a multilevel logistic regression to model the ranks. This is very easily done in STATA, using the procedure for Rank ordered Logistic Regression.

Dear Kanika,

Ranked data is not independent. A better test would be Friedman's ANOVA.

If you want to test the interaction between gender and the 10 reasons with a nonparametric test, you can use the nparLD or the npIntFactRep R packages. Within the nparLD package (see attachment), chose the f1.ld.f1 design. For the nparLD package the input data should be in 'long' format, for the npIntFactRep package, they should be in 'wide' format. This latter package only calculates if the interaction is significant. There is also a SAS macro for f1.ld.f1; see attachment.

With the Friedman test you can only test the effect of reasons, not of gender nor the interaction. (By the way, the aligned ranks Friedman test has more power than the regular Friedman test. In SAS, the data (in long format) are first standardized by subject, then ranked and then the Friedman test is run on these (aligned, i.e. standardized by subject) ranks:

proc standard mean=0;

by subject;

var outcome;

proc rank;

var outcome;

proc freq;

tables subject*type*outcome / noprint cmh2;

run;)