Simulated correlations with different magnitudes and noise and recovered them. - partial pooling is better for all but tau/ndt parameter # alpha/boundary ```r # correlations for panels below > p_grid[, .(r = round(cor(sim_cor, recovered_cor_nopool), 4)), keyby = .(met, e)] met e r 1: pearson 1 0.9993 2: pearson 5 0.9931 3: pearson 9 0.9894 4: pearson 13 0.9391 5: pearson 17 0.9535 6: spearman 1 0.9996 7: spearman 5 0.9907 8: spearman 9 0.9874 9: spearman 13 0.9795 10: spearman 17 0.9686 > p_grid[, .(r = round(cor(sim_cor, recovered_cor_pool), 4)), keyby = .(met, e)] met e r 1: pearson 1 0.9998 2: pearson 5 0.9962 3: pearson 9 0.9960 4: pearson 13 0.9697 5: pearson 17 0.9855 6: spearman 1 0.9995 7: spearman 5 0.9931 8: spearman 9 0.9930 9: spearman 13 0.9880 10: spearman 17 0.9766 # larger abs deviation from true value t = 8.6945, df = 209, p-value = 1.029e-15 # larger SEs for no-pooling t = 5.0574, df = 104, p-value = 1.835e-06 ``` ![[alpha-bound 1.png]] # tau/ndt ```r # correlations for panels below > p_grid[, .(r = round(cor(sim_cor, recovered_cor_nopool), 4)), keyby = .(met, e)] met e r 1: pearson 1 1.0000 2: pearson 5 0.9996 3: pearson 9 0.9988 4: pearson 13 0.9952 5: pearson 17 0.9970 6: spearman 1 1.0000 7: spearman 5 0.9994 8: spearman 9 0.9988 9: spearman 13 0.9969 10: spearman 17 0.9960 > p_grid[, .(r = round(cor(sim_cor, recovered_cor_pool), 4)), keyby = .(met, e)] met e r 1: pearson 1 1.0000 2: pearson 5 0.9996 3: pearson 9 0.9989 4: pearson 13 0.9940 5: pearson 17 0.9973 6: spearman 1 0.9999 7: spearman 5 0.9993 8: spearman 9 0.9990 9: spearman 13 0.9965 10: spearman 17 0.9959 # no diff in deviation t = -0.007922, df = 209, p-value = 0.9937 # no difference in SE t = 0.86532, df = 104, p-value = 0.3889 ``` ![[tau-ndt 1.png]] # beta/bias ```r # correlations for panels below > p_grid[, .(r = round(cor(sim_cor, recovered_cor_nopool), 4)), keyby = .(met, e)] met e r 1: pearson 1 0.9989 2: pearson 2 0.9967 3: pearson 3 0.9912 4: pearson 4 0.9821 5: pearson 5 0.9866 6: spearman 1 0.9989 7: spearman 2 0.9953 8: spearman 3 0.9900 9: spearman 4 0.9845 10: spearman 5 0.9827 > p_grid[, .(r = round(cor(sim_cor, recovered_cor_pool), 4)), keyby = .(met, e)] met e r 1: pearson 1 0.9991 2: pearson 2 0.9973 3: pearson 3 0.9907 4: pearson 4 0.9878 5: pearson 5 0.9905 6: spearman 1 0.9988 7: spearman 2 0.9951 8: spearman 3 0.9917 9: spearman 4 0.9886 10: spearman 5 0.9825 # larger abs deviation from true value t = 4.5721, df = 309, p-value = 6.995e-06 # larger SEs for no-pooling t = 3.9775, df = 154, p-value = 0.0001069 ``` ![[beta-bias 1.png]] # delta/drift ```r # correlations for panels below > p_grid[, .(r = round(cor(sim_cor, recovered_cor_nopool), 4)), keyby = .(met, e)] met e r 1: pearson 1 0.8723 2: pearson 3 0.5252 3: pearson 5 0.3604 4: pearson 7 0.3899 5: pearson 9 0.2182 6: spearman 1 0.9497 7: spearman 3 0.8066 8: spearman 5 0.7972 9: spearman 7 0.5895 10: spearman 9 0.4286 > p_grid[, .(r = round(cor(sim_cor, recovered_cor_pool), 4)), keyby = .(met, e)] met e r 1: pearson 1 0.9631 2: pearson 3 0.8403 3: pearson 5 0.4730 4: pearson 7 0.5479 5: pearson 9 0.5882 6: spearman 1 0.9629 7: spearman 3 0.8730 8: spearman 5 0.7805 9: spearman 7 0.6522 10: spearman 9 0.5356 # larger abs deviation from true value t = 6.5905, df = 409, p-value = 1.354e-10 # larger SEs for no-pooling t = 4.6021, df = 204, p-value = 7.342e-06 ``` ![[delta-drift 2.png]]