Hi all,
My respondents were 75% males (233) and 25% females (76). and gender is the moderating variable. based on the findings, gender does not moderate the relationship. and when I ran the t-test, there was no significant difference between males and females on the continuous measure. Do these findings (no moderation and non-significant t-test) due to the unequal size of gender groups?
Thank you.
No. The exception is a really small sample. Some numbers of sample are too little for counting correlations, for instance. If your 25% is 50 and more female participants - t-test result may be considered as true. Besides, you should pay attention to the response rate.
Dear Arnida
It might deppend on the population under study. eg: If you are selecting a random sample of electors for an political election, usually the number of males and females in this population is very similar, then the lack of female answer could be an indication of different response disposition for each sex. You should research the real motivations of not response to avoid a bias of selection/response. If you population is that of the pupils of an engenier career, the number of males and females may be in the same proportion in the population and then most probabilistic samples would show this situation.
Hello Arnida
my 2 cent (beware, these are speculations)
1) As any reduction in variance of an IV, an imbalance should increase the standard error, thus, you reduce power
2) As gender is exogeneous, you don't get a (selection) bias. Selection bias concerns a systematic non-response of the dependent variable
HTH
--Holger
It is unclear how your model really looks like, since you are talking about moderation and then t-test. Was it a simple t-test without considering the X variable?
Besides that, I would like to add to Holgers comment (and contrary to Marianas comment) that even in larger samples power may be drastically affected by unequal sample sizes. Just take a simple between t-test (two-tailed) with N=309 (as in your sample) with (nearly) equal samples n1 = 155 and n2 = 154. For an alpha = 0.05 and an effect size of d = 0.3 you would have a power of .75. With your samples n1= 233 and n2 = 76, this would reduce to .64.
- Badges
- Science topic