Skip to main content
. Author manuscript; available in PMC: 2020 Apr 29.
Published in final edited form as: J Neuroimmune Pharmacol. 2018 Oct 1;13(4):532–540. doi: 10.1007/s11481-018-9811-8

Figure 4.

Figure 4.

Logistic regression models built with LASSO selected predictors. The sPGSs (solid lines) and tPGSs (dashed lines) were tested separately. The MGS+SSCCS data were used as training data, and the independent CATIE sample was used as testing data. The sPGS model had 19 predictors and the tPGS model had 14 predictors. The sPGS model outperformed tPGS model in both training (red lines) and testing (green lines), but it had an issue of overfitting in the training data.