A Web-Disseminated Self-Help and Peer Support Program Could Fill Gaps in Mental Health Care: Lessons From a Consumer Survey.

Retaining “marginally significant” covariates in the final model (ie, using a cutoff of P <.10 as opposed to P <.05) constitutes a more conservative test of the independent predictive ability of each variable, while eliminating those that did not approach significance in earlier models reduces noise and increases parsimony.

also in 26,082 other papers

In the first hierarchical logistic regression predicting whether participants provided an email address to request more information, blocks including gender, race, symptoms, and hypothetical treatment use significantly improved model fit; no other blocks approached significance.

also in 8,237 other papers