Welcome to the Biostat done for you podcast with Nouran Hamza where I’ll help you to make the most of your clinical research and contribute to informed evidence based decisions
I was just talking to one of our clients, he asked me why there are no details on the normality testing for his regression model !
and I believe you might (for example) run a clinical research to find out if blood pressure rises linearly with age).
And you would choose to run You might run clinical research to find out if blood pressure rises linearly with age).
And you would choose to run a linear regression model, you will find some blogs telling you to test the normality of your observed data
surprise surprise , the linear regression assumptions are about the residuals , the variation between the predicted and observed values.
The easiest way to understand if your model fits is by understanding the diagnostic plots ….. ,
The first one is the Residuals vs Fitted plot, it will be pretty good if you find A horizontal line without any distinctive patterns, because that indicates a linear relationship.
Also the Q-Q plot that shows whether the residuals are normally distributed. If the residual points follow this straight, dashed line, this is a huge success.
Spread-Location (or scale-Location). Used to check the homogeneity of variance of the residuals (homoscedasticity). Horizontal line with equally spread points is a good indication of homoscedasticity.
you will find also the Residuals vs Leverage plot, which is Used to identify influential cases, There may be extreme values that will influence the regression results when included or excluded from the analysis, don’t ever remove any data point blindly
Tell hear again from you, have a nice day