Data and Models in Psychology

statistical practice

OLS assumptions

robust methods

Sladekova, M., Jakeman, H., Field, A. P.

Stage: Data analysis underway

Continuing on from the investigation of the State of Statistical Practice in Psychology, this study has two main goals. First, we want to find out about the data and models typically found in in psychology - what does the typical distribution of residuals look like, and how common are outliers?

Second, we want to estimate the potential impact these characteristics may have on the OLS estimates and related hypothesis tests. To do so, we’ll investigate the extent to which OLS estimates change when the models are re-fitted with alternative robust estimators that don’t rely on the same statistical assumptions.

To date, we’ve re-analysed ~600 models from 120 papers (including published papers and pre-prints). We’re currently in the process of extracting different assumption metrics out of the models, which will then inform a follow-up simulation study.