Lassologit

## Logistic lasso #

lassologit is intended for classification tasks with binary outcomes. lassologit maximizes the penalized log-likelihood:

$\frac{1}{N} \sum_{i=1}^N \left\{y_i (\beta_0 + \boldsymbol{x}_i'\boldsymbol{\beta}) - \log\left(1+e^{(\beta_0+\boldsymbol{x}_i'\boldsymbol{\beta})}\right)\right\} - \frac{\lambda}{N} ||\boldsymbol{\beta}||_1$

where $$y_i$$ is the binary outcome variable and $$\boldsymbol{x}_i$$ is the vector of predictors. $$\boldsymbol{\beta}$$ is the vector of parameters to be estimated. The last term in the objective function imposes a penalty on the absolute size of $$\boldsymbol{\beta}$$ . The intercept $$\beta_0$$ is (by default) not penalized.

lassologit implements the coordinate descent algorithm of Friedman, Hastie & Tibshirani (2010, Section 3). For further speed improvements, we also utilize the strong rule proposed in Tibshirani et al. (2012).

Like lassopack, lassologit consists of three programs which correspond to three approaches for selecting the tuning parameter $$\lambda$$ :

• The base program lassologit allows to select the tuning parameter as the value of $$\lambda$$ that minimizes either $$AIC$$ , $$BIC$$ , $$AIC_c$$ or $$EBIC$$ .
• cvlassologit supports $$K$$ -fold cross-validation. $$\lambda$$ may be selected as the value that minimizes the estimated deviance or miss-classification rate.
• rlassologit implements theory-driven penalization for the logistic lasso (see e.g. Belloni, Chernozhukov & Wei, 2016).
Installation
Lassologit has been integrated into lassopack after the first release. To get the latest lassologit version, simply install lassopack.