If your underlying question is "Are small groups more or less likely to help", then you want to produce a model that has output like "chance of helping" in terms of inputs relating to the dependent variables. If you expect a possibility of a sex (M/F) effect, you should include it as a covariate.
The most obvious choice would be a logistic regression model, with DVs of sex and group size.
It might also be possible to do it using chi-square, but if you're allowing for both DVs, you'll have a 2x2x$k$ table, where $k$ is the number of group sizes. One advantage of the logistic regression is it's easier to model group size as continuous, and if there are more than two group sizes, you can more easily test for some form of ordering (i.e. deal with effects like like "as the group size increases, people are less likely to help").