why is it a good thing to run both a bivariate and a multivariate analysis when determining the relationship between two variables? what changes happen when other variables are introduced into the models?