Altair slc chapter I optimum binning in preparation for logistic regression

Question

Too long to post on a listserve, see github

github

https://github.com/rogerjdeangelis/utl-altair-slc-chapter-I-optimum-binning-in-preparation-for-logistic-regression

Key output reports

Binned character covariates excel report

https://github.com/rogerjdeangelis/utl-altair-slc-chapter-I-optimum-binning-in-preparation-for-logistic-regression/blob/main/lgs_MgmChrCut.xlsx

Binned numeric covariates excel report

https://github.com/rogerjdeangelis/utl-altair-slc-chapter-I-optimum-binning-in-preparation-for-logistic-regression/blob/main/lgs_MgmNumCut.xlsx

SECTIONS

I   ANALYSIS OVERVIEW

II   CAMPAIGN STRATEGY

III   KEY OUTPUT SAMPLE OF BINNING ONE COVARIATE

IV   INTERPRETING BINNED DATA

V   INPUT

VI   OUTPUT FIVE TABLES

VII   THEORY

VIII   METHOD  (program

1   Contents of raw input data

2   VALIDATE AND VERIFY RAW DATA. FIX ISSUES WITH RAW DATA

3   One missing vale

Drop one to one variable. Very low caedinality variables.

Convert the many missing variable formats to just '?'

4   Optimize variable length to the longest in the data

5   Create Holdout and Training Tables

6.  Bin the character variables in groups with common odds ratios

Create normalized (long and skinny binned data wirh just 55 variables)

Create denormalized (wide binned table with 30 variables)

7   Create excel character data summary bin report with chi square and  Mantel–Haenszel Stats

8   Bin the numeric variables in groups with common odds ratios

Create normalized (long and skinny binned table wirh just 55 variables)

Create denormalized (wide binned table with 38 variables)

9   Create excel numeric data summary bin report with chi square and  Mantel–Haenszel Stats

10   Join raw training table, numeric binned data with character binned data for analys

IX    CONTENTS OF FUTURE CHAPTERS

Logistic diagnosis and related reports

Fitting logistic regression on training data

lgs_mgmFinalLogisticDiag   Logistic Model (Diagnostics)

lgs_MgmGainsChart          Gains Chart

lgs_mgmTopChiValues        Most Influential Variables

lgs_mgmTopIndexValues      Highest Response Variables

lgs_MgmTopTen              List of Top 12 scores

lgs_MgmVenn n              Comparison of covariate contribution to top pcile

Final pdf presentation of results (slidedeck)