Hi there

Thanks for your Information package.

I have been using the create_infotables and plot_infotables function but I was wondering about the sign of the WOE value. Using the German credit data, variable checkingstatus1 level A11 for example, I manually calculate the WOE as -0.818 whereas it is calculated as 0.818 in create_infotables function. As such the plot_infotables plots the positive 0.818. My calculation of the IV is the same ie 0.206.

See below my example:

Load german credit data

url <- ("http://www.biz.uiowa.edu/faculty/jledolter/DataMining/germancredit.csv")
creditdata <- readr::read_csv(url)

Use the CrossTable function from gmodels package on default and checkingstatus1

gmodels::CrossTable(creditdata$Default,
creditdata$checkingstatus1, prop.r= TRUE, prop.c =FALSE, prop.t=FALSE,
chisq = TRUE)

Manually calculate WOE for A11

percentgood <- 0.1985714
percentbad <- 0.45
odds <- percentgood/percentbad
woe <- log(odds)

Create a WOE table with IV values

IV <- create_infotables(data=creditdata,
y="default")
plot_infotables(IV,"checkingstatus1", show_values=TRUE)

suuuuuper blog
https://multithreaded.stitchfix.com/blog/2015/08/13/weight-of-evidence/
and git repo
but only can you share it in python code?
or may be you know some exteranl link for python ?

Hi,

Just wondering if the WOE & IV handles factor columns differently than if they were some type of integer column/id? I have seen some different variable outcomes when using create_infotables()
when having my character columns converted to factors versus when I convert them to numeric id's.