Using Multivariate Tests as Univariate Tests

You may already have figured this out by now, but you can use multivariate analyses as univariate analyses. All you need to do is put a single predictor variable in your model, and voilà – a univariate analysis.

The result you get will be the same as (or at least, very similar to) the result you would get from the equivalent univariate analysis in terms of the p-value, coefficients, OR, and so on.

Some researchers prefer to use multivariate analysis tools for their univariate analysis before going on to do a full multivariate analysis with the same tools. The advantage to this is that they get consistency in the tools they use.

I prefer not to do this, though. I like using different tools for univariate and multivariate analysis mostly because the univariate tools give you a better 'feel' for your data, but also because I feel more confident in the results if they concur and have been computed by different methods.