Vcemway: A one-stop solution for robust inference with multiway clustering

Ariel Gu, Hong Il Yoo

Research output: Contribution to journalArticlepeer-review

13 Citations (Scopus)

Abstract

Most Stata commands allow cluster(varname) or vce(cluster clustvar) as an option, popularizing the use of standard errors that are robust to oneway clustering. For adjusting standard errors for multiway clustering, there is no solution that is as widely applicable. While several community-contributed packages support multiway clustering, each package is compatible only with a subset of models that Stata’s ever-expanding library of commands allows the researcher to fit. We introduce a command, vcemway, that provides a one-stop solution for multiway clustering. vcemway works with any estimation command that allows cluster(varname) as an option, and it adjusts standard errors, individual significance statistics, and confidence intervals in output tables for multiway clustering in specified dimensions. The covariance matrix used in making this adjustment is stored in e(V), meaning that any subsequent call to postestimation commands that use e(V) as input (for example, test and margins) will also produce results that are robust to multiway clustering.

Original languageEnglish
Pages (from-to)900-912
Number of pages13
JournalStata Journal
Volume19
Issue number4
DOIs
Publication statusPublished - 1 Dec 2019

Keywords

  • boottest
  • cmgreg
  • ivreg2
  • multiway clustering
  • reghdfe
  • st0582
  • two-way clustering
  • vcemway

Cite this