Regression when the Normality Assumption is Violated

If there’s one caveat that most of us remember about least squares regression, it’s this: Regression assumes normality. More specifically, least squares regression assumes that the distribution of Y given X is normal, or equivalently, that the distribution of residuals is normal. But what if our data violate this assumption (e.g., our response distribution is highly skewed or consists of only integer values), potentially causing us to make incorrect inferences from the results? Thankfully, there are techniques to easily handle this all-too-common scenario.

This webinar, intended for academic researchers, will teach you concepts and techniques for conducting linear regression with non-normal response data, including how to implement these techniques in the JMP Student Edition, a no-cost, full-featured version of JMP exclusively for academic use. Topics include:

• Assessing normality

• Transforming response variables

• Generalized linear models (GLMs) that use a more appropriate response distribution

• Selecting the best fitting distribution for your GLM

Get JMP software free for academic use at jmp.com/student