pt_winsor: Winsorization of outliers

In PBCAR/PThelper: Systematic Cleaning, Processing, and Summarizing of Behavioral Economic Purchase Task Data

Winsorization of outliers

Description

This function helps users to manage outliers at the indicator level by using winsorization techniques. See details for the different options.

Usage

pt_winsor(pt, id_var, z_val = 3.99, option = 3, w_var = NULL, delta = NULL)

Arguments

pt	A data frame consisting of the 'id_var' and relevant purchase task variables.
id_var	The name of the unique identifier (ID) as identified in the data frame.
z_val	The absolute z-score value to define outliers. The default is |3.99|, which will remove those exceeding a z-score of |3.99|.
option	The winsorization option, one of c(1,2,3). The default outlier management technique is option 3.
w_var	The name of the variable to winsorize.
delta	The constant used in winsorization options 2 and 3. The delta must be defined by the user, as the optimal value will vary depending on the indicator. For elasticity, a small value of 0.001 is recommended.

Details

There are 3 winsorization options:

i) Option 1 replaces all outliers with the theoretical value associated at the defined z-score threshold;

ii) Option 2 replaces all outliers with the observed minimum/ maximum non-outlying value plus (or minus) a small constant (delta);

iii) Option 3 replaces each outlier with the observed minimum/ maximum non-outlying value plus a small constant (delta) to retain order.

Value

A list consisting of two data frames: "data" which consists of the 'id_var' and 'pt' including the winsorized value(s); and "wins_table" which provides details on which value(s) by 'id_var' were winsorized (values before and after provided).

Examples

### --- Example Data
pt <- data.frame("ID" = c(1:36),
"Intensity" = c(10,12,15,0,99,11,7,6,12,7,8,10,5,6,10,0,3,
                7,5,0,2,3,5,6,10,15,12,7,0,9,0,6,7,8,4,5))

### --- Function Example
pt2 <- pt_winsor(pt, id_var = "ID", w_var = "Intensity", delta = 1)

PBCAR/PThelper documentation built on May 13, 2024, 3:45 p.m.