Figure 3. Regression models estimate elevation’s association while capturing known risk factors.

(A) Summary of the predictors included in the optimal best subset model for each cancer (see Table S1 for more detail). (B) Summary of the models produced by lasso regression, displaying characteristic coefficient shrinkage. Both regression techniques produced similar sets of models that were sensible for lung, breast, and colorectal cancer. Elevation displayed a strong and consistent negative coefficient in lung cancer models.