dev_logit: Deviance from predicted values of a logit model

In kuriwaki/sparseregMRP: MRP Estimates from sparsereg Regressions

Description

Deviance from predicted values of a logit model

Usage

dev_logit(y, xb = NULL, phat = NULL, w = 1)

Arguments

y	vector of the outcome that was regressed on, or the true values to predict
xb	estimates from the logit model, before the link function is applied.
phat	estimates from the logit model, after link function is applied. Only either xb or phat is necessary.
w	survey weights if appplicable

Details

The deviance without weights, dev(\hat p) for target y is computed as

-2 \times ∑_{i=1}^n y_i \log(\hat{p_i}) + (1 - y_i) \log(1 - \hat{p_i})

where

\hat{p} = \frac{\exp{(Xβ)}}{\exp{(Xβ)} + 1}

.

With weights, this simply becomes

-2 \times ∑_{i=1}^n w_i y_i \log(\hat{p_i}) + w_i (1 - y_i) \log(1 - \hat{p_i})

kuriwaki/sparseregMRP documentation built on March 20, 2020, 9:58 p.m.