Fig. 2
From: Consequences of ignoring clustering in linear regression
Difference from “true” value of 1 of regression coefficients estimated from RI and OLS models (\({\beta }_{1}^{RI}\) and \({\beta }_{1}^{OLS}\)) plotted against dispersion (expressed as SD) of the mean value/prevalence of \({x}_{ij}\) across the clusters within each sample. Results for different levels of intraclass correlation coefficient are distinguished by shades of grey as indicated in the legend. A Continuous \({x}_{ij}\). B Binary \({x}_{ij}\)