Skip to main content

Table 1 Coverage (%) of “true” effect \({\beta }_{1}\) = 1 by 95% confidence intervals derived from the RI and OLS models according to fifths of the distribution of dispersion (expressed as SD) of the cluster means of a continuous \({x}_{ij}\) within samples

From: Consequences of ignoring clustering in linear regression

  Lowest fifth of distribution 2nd fifth of distribution 3rd fifth of distribution 4th fifth of distribution Highest fifth of distribution Total
RI OLS RI OLS RI OLS RI OLS RI OLS RI OLS
0.001 95.04 94.37 95.00 94.09 95.33 94.10 95.15 93.99 94.92 93.84 95.08 94.08
0.003 95.14 93.14 95.37 92.33 95.20 91.83 95.53 92.25 95.42 91.99 95.33 92.30
0.01 94.99 88.47 94.64 83.95 94.75 83.65 94.74 83.72 94.90 84.07 94.80 84.75
0.03 94.59 76.21 95.11 68.79 94.80 67.62 95.06 67.57 94.74 67.15 94.87 69.39
0.1 94.68 59.58 94.80 45.45 94.86 44.83 94.39 44.73 95.04 44.37 94.76 47.80
0.3 94.84 41.32 94.53 28.06 94.95 28.24 94.64 26.98 94.79 27.41 94.75 30.36