multivariate comparaions - p value correction & post hoc parameter choice

uchicago-computation-workshop / abdullah_almaatouq

Repository for Abdullah Almaatouq's presentation at the CSS Workshop (4/11/2019)

0 stars 0 forks source link

multivariate comparaions - p value correction & post hoc parameter choice #30

Open liaoandi opened 5 years ago

liaoandi commented 5 years ago

Hi, thanks for the presentation! It is interesting. However, I have a question about statistical analysis. As you used multiple groups, multiple conditions and multiple phases to compare, maybe it is essential to use p-value correction methods? Plus, when choosing the top-k participants, similar correction techniques might be necessary in post-hoc parameter choice.
Thanks!

amaatouq commented 5 years ago

yes, all p-values on individual-level outcomes are based on White cluster-robust standard errors to account for correlated error terms within each group. Also, all analyses/hypothesis/comparisons were pre-registered.