Multivariate statistics for continuous outcomes are used to adjust for the demographic, clinical, and confounding effects that exist between predictor and continuous outcome variables. The type of multivariate statistic used for a continuous outcome depends upon the nature of the research question and research design. With multivariate statistics, the continuous outcome variable needs to be normally distributed to be used in most of the analyses below. Non-normal distributions that occur naturally are considered count outcomes. Different multivariate statistics are needed for these types of continuous outcomes. Finally, multivariate statistics can be used to test multiple outcomes concurrently so as to decrease experimentwise error rates.
Researchers are testing how an outcome varies across multiple categorical predictor variables.
Researchers are testing how multiple outcome variables vary across time or within-subjects.
Researchers are testing how different groups change across time.
Researchers are adjusting the value of a continuous outcome across independent levels of a categorical predictor variable.
Researchers are establishing the associations amongst multiple predictor variables and a continuous outcome variable.
Researchers want to account for increased Type I error rates when testing multiple associations and have multiple continuous outcomes of interest.
Researchers want to predict for a count outcome such as with Poisson Regression or Negative Binomial Regression.
