3
$\begingroup$

Are there commonly accepted ways to visualize the results of a multivariate regression for a non-quantitative audience? In particular, I'm asking how one should present data on coefficients and T statistics (or p-values) for a regression with around 5 independent variables.

$\endgroup$
2
$\begingroup$

I personally like dotcharts of standardized regression coefficients, possibly with standard error bars to denote uncertainty. Make sure to standardize coefficients (and SEs!) appropriately so they "mean" something to your non-quantitative audience: "As you see, an increase of 1 unit in Z is associated with an increase of 0.3 units in X."

In R (without standardization):

set.seed(1)
foo <- data.frame(X=rnorm(30),Y=rnorm(30),Z=rnorm(30))
model <- lm(X~Y+Z,foo)

coefs <- coefficients(model)
std.errs <- summary(model)$coefficients[,2]

dotchart(coefs,pch=19,xlim=range(c(coefs+std.errs,coefs-std.errs)))
lines(rbind(coefs+std.errs,coefs-std.errs,NA),rbind(1:3,1:3,NA))
abline(v=0,lty=2)

regression visualization

$\endgroup$

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.