I hope this help you: http://pbpython.com/categorical-encoding.html

If you just add a dummy variable (0/1 coding), then you are only examining a direct effect of gender. To get a moderated analysis, you would also need to form 4 interaction terms (gender x IV for each of the 4 IVs).

Take a look at the Sage "little green book" by Hardy, Regression with Dummy Variables.

Thank you both for your replies.

David, I do expect that gender will have an impact on each of the IVs so I suppose that means I do need a moderated analysis.

Would it be normal to do both i.e. add a dummy variable and run a multiple regression to test for a direct effect of gender on the DV and then run a dichotomous moderated multiple linear regression analysis to see if there's interaction between gender and each IV?

Or can you read a direct effect of gender on the DV from the dichotomous moderated multiple linear regression test ?

Finally, would SEM / CFA be suitable for my purposes?

The standard approach is to begin with an equation that includes just the dummy variable for gender, and then run a further equation that adds the 4 variables for the interaction effects. The difference in R-sq between these two equations will give you a "global" test of whether the regressions are the same for men and women, while the significance of the 4 individual coefficients will tell you which specific variables differ for men and women.

I am not sure what you mean by a "dichotomous moderated multiple linear regression analysis," but including the 4 interaction effects in a single equation along with the dummy variable is mathematically equivalent to running the two regressions separately (i.e., one regression for males and one for females). The advantage of using the 4 interaction effects is that you get a significance test for each.

You could use SEM but that should give you a result that is identical to the one from using the four interaction effects.

Thank you David for your answer. Embarrassingly, I feel it's somewhat above my own level of understanding.

Is it possible to have 2 categorical IVs as moderators along with 3 continuous IVs and 1 continuous DV? If so is that just a moderated multiple regression?

Having two moderators operating at the same time creates another whole level of complexity. For example, if each moderator has 2 categories, you have to create a 2x2 set of interactions to cover the full set of possibilities. This represents the possibility that each moderator moderates the effect of the other one.

I wish there were an easier way to do this, but I can't think of one. I recommend that you read up on interaction effects in regression. Hardy's "little green book" from Sage on dummy variables is a good place to start. For a more complete, but more advanced, book, Jaccard on Interaction Effects in Multiple Regression (in the same series), would be a good next step.

Thank you David :)

I may just have to choose between the 2 moderators then. It's only for a proposal so I won't be doing any statistical tests, I just need to figure out which one would hypothetically be the most workable to examine the relationship between 4IVs and 1DV but within 2 groups (male vs female and internet users vs non internet users).

Having 2 moderators would translate into the following hypothesis for an interaction effect: the effect of being an internet user or not differs between men and women.

The combination of an IV and the 2 x 2 interaction could take on a number of forms, but the simplest version would be something like: the effect of the IV among women who are internet users is distinctly different from its effects in the other 3 subgroups.

Thanks David

The simplest approach is taken by Paul Jose website on moderation. Type it in google.