gender differences


I have a daily information for years about a dependent variable that measures the degree of risk taken by men and women in their decision making ( we assigned values of 1 for the decision with the highest degree of risk aversion , then 2 for the riskier decision , and i moved progressively up to 9 to categorize the risk of decisions). I want to test the effect of gender differences on this decision making ; besides gender I have information about age, marital status,occupation and income. I am wondering what is the best design to test for the gender differences?


New Member
If the dependent variable is continuous (that is higher scores indicate riskier decisions), then a t-test is most appropriate
I meant what about the regression? what kind of regression I should run when the dependent variable can take a value from 1 to 9 (a rank). I assume i can use the T-test of the coefficient of the gender to test if gender matters but does it make sense to run an OLS in this case?


Super Moderator
Whether you can use an OLS kinda depends on whether you can justify treating your 9-value rating scale as an interval-level continuous variable. Strictly speaking it isn't, but it could be justifiable - you at least have a large-ish number of possible values. Would you be able to defend saying, for instance, that the difference between a score of 1 and a score of 3 is the same as the difference between a score of 7 and a score of 9?

The distribution of scores on the scale (normal or otherwise) is also relevant to the question of whether an OLS regression is ok.

Anyway, if you do treat the scale as interval/continuous, t-test for 2 independent samples or a Mann-Whitney test would do fine for bivariate differences (depending on the distribution of the scale scores), or you could use a hierarchal linear regression to check whether the gender difference is significant after your other variables are controlled for.
And i wouldnt refer to "gender" differences. Rather, "***" refers to biological makeup (whereas gender is considered more "which group you identify with").


Super Moderator
not sure why that is bleeped out. starts with s, ends with x.

Haha! Puritanical old talkstats... you make an excellent point though, gender ≠ s... although the talkstats block is a tres pertinent example of how researchers end up using 'gender' - it means the same thing and is more polite, right? :p Sigh!