Fig. 8.

Bias (left panels) and standard error (i.e. SE, right panels) of log transformed odds ratio estimations for different effect sizes of EC. Each line indicated one model. Note: model 1, unconditional logistic regression model without adjusting for C for non-matched case-control designs; model 2, unconditional logistic regression model with adjusting for C for non-matched case-control designs; model 3, unconditional logistic regression model without adjusting for C for matched case-control designs; model 4, unconditional logistic regression model with adjusting for C for matched case-control designs; model 5, conditional logistic regression model with adjusting for C for matched case-control designs