3
$\begingroup$

What is the best way to visually present effect sizes for multiple regression predictor variables? My research deals with home prices and the factors that influence price. I have transaction data showing sales price and physical/location attributes of the property. I am usually presenting to audiences with low stats IQ and feel that charts would better understood than just plain regression output.

A scatter plot (predictor variable vs price) usually shows the general relationship. Is there a way to create an equivalent chart, but showing effect size (in the context of the full model?). And also handle quadratic terms or other predictor variable transformations?

I am just beginning to learn R, and currently Use Minitab

$\endgroup$
1
$\begingroup$

A scatterplot would only show 1 predictor at a time, or 2 if you do it as a 3D or density plot. By effect size in the context of the full MR model you, presumably, mean adjusted R2 or relative importance of each predictor. Which does not show actual data but can show many predictors at once: e.g. explore relaimpo in R, although you can just output bootstrapped measures of relative importance and graph them any other way you like.

Make sure you are not over-fitting with polynomial terms (which are not predictor variable transformations but a model term, whereas transformations should be chosen prior to analyses, based on data distibution and your underlying hypothesis).

$\endgroup$
0
$\begingroup$

If you want to keep the layout of a scatterplot and want to show the effect size, I would use a bubble plot to show this information.

Have a look at this tutorial by Nathan Yau on how to make nice bubble plots in R.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.