Using the Least Absolute Shrinkage and Selection Operator for Variable Selection and Model Building

• In Event: Missing Data, Multicollinearity, and Variable Selection in Regression

Abstract

The least absolute shrinkage and selection operator (lasso) is a regression analysis technique that performs automatic variable selection while fitting linear models. Lasso overcomes several disadvantages of traditional regression by preventing model overfitting and allowing the analysis of high-dimensional data. The variable selection feature of the lasso allows data to include an unlimited number of independent variables; despite this advantage, lasso is not frequently cited in educational research. The current paper seeks to introduce the lasso regression technique. Two demonstrations are shared to illustrate the lasso's use on empirical and simulated datasets in R. All of the demonstration code and data are included. Variations of the lasso algorithm for different data scenarios are briefly discussed.

Authors

• Christina Witta Gillespie, Valencia College

• Benjamin Witta, Texas A&M University - Corpus Christi