How can I get cluster-robust errors (for generating confidence intervals) on a Pearson correlation coefficient?

I know that I can generate cluster-robust errors for a coefficient in a linear model. And I know that square-root of the r-squared in the linear model is equivalent to the Pearson correlation. But I don't know how to get an appropriate standard error on the Pearson correlation.

Note: this is how i'm accounting for cluster-robust errors


Standard error alone cannot be used to generate a confidence interval for a Pearson correlation coefficient. One way to generate confidence intervals is to do a 'percentile bootstrap' (see here). Instead of bootstrapping a standard error, use the bootstrap process to generate a distribution of correlation coefficients. Then, identify the bounds the inner (eg) 95% of the data to generate a (eg) 95% confidence interval.

To account for clustered data, use a clustered bootstrap: instead of sampling-with-replacement N individuals for each bootstrap estimate, sample entire clusters. (see a practitioners guide to cluster robust inference)

| cite | improve this answer | |

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.