stepwise_regression: Forward entry stepwise regression

In jeromyanglim/personalityfacets: Package for estimating population r-square change

Description

This function performs forward entry stepwise regression.

Usage

stepwise_regression(dv, ivs, data, alpha_in = 0.05)

Arguments

dv	character string representing name of dependent variable in data
ivs	character vector representing predictor variables in data
data	data.frame
alpha_in	maximum p-value permitted of r-square change of non-included predictor to allow inclusion of predictor

Details

The model starts with no predictors and determines which predictor results in the largest increase to r-squared. If this predictor has a p-value less than alpha_in then the predictor is retained. If not then the predictor is excluded and the process stops.

Then all remaining predictors are examined. The predictor with the largest squared semi-partial correlation adjusting for predictors already in the model is further examined. If this predictor has a p-value less than alpha_in (e.g., .05) then the predictor is retained. If it is not significant, then the stepwise procedure ends and the model is returned. If it is significant, thent the procedure is iterated again.

Note that this function is included mostly for comparion purposes. We do not advocate the use of stepwise regression for the purpose of evaluating the incremental prediction of facets over factors.

Value

The model fit of class lm returned from the final model. The order of predictors corresponds to the order that they entered the model.

Examples

data(facets_data); data(facets_meta)

## Example with the Big 5 as predictors:
stepwise_regression('swl', c("ipip_neuroticism", "ipip_extraversion",
     "ipip_openness", "ipip_agreeableness", "ipip_conscientiousness"),
     facets_data, alpha_in=.05)

## Example using 30 facets as predictors
## (illustrating a more stringent alpha_in):
stepwise_regression('swl', facets_meta$ipip_facets,
     facets_data, alpha_in=.001)

jeromyanglim/personalityfacets documentation built on Dec. 23, 2019, 7:31 p.m.