Gains Chart Analysis for Scoring Model Performance Evaluation
School
University of California, Berkeley**We aren't endorsed by this school
Course
UGBA 10
Subject
Economics
Date
Dec 11, 2024
Pages
5
Uploaded by MPi2022
This Spreadsheet illustrates a GAINS CHART to evaluate performance of a scoring modelFor convenience of demonstration the Gains Chart is being built in the estimation sample. Ideally yDATAGAINS CHART IndividualPurchase# exposed%Exposed#Hit%Hit111.00410.033310.0833210.39720.066720.166730-0.16530.100020.1667400.23740.133320.1667500.71950.166720.166760-0.12860.200020.1667700.31770.233320.1667810.97480.266730.2500910.68290.300040.33331010.587100.333350.41671110.602110.366760.50001210.806120.400070.58331300.178130.433370.58331400.208140.466770.58331500.054150.500070.5833160-0.041160.533370.5833170-0.347170.566770.58331800.565180.600070.58331910.806190.633380.66672000.142200.666780.66672100.339210.700080.6667220-0.055220.733380.6667230-0.004230.766780.66672410.543240.800090.75002510.646250.8333100.83332610.806260.8667110.91672700.302270.9000110.91672800.186280.9333110.91672900.616290.9667110.91673011.025301.0000121.000012To assess the model, sort the data (yellow and green columns) in descending order of Score (coluMean Random0.5167Score (=predicted Y)
you would do this in a hold-out sample0.03330.06670.10000.13330.16670.20000.23330.26670.30000.33330.36670.40000.43330.46670.50000.53330.56670.60000.63330.66670.70000.73330.76670.80000.83330.86670.90000.93330.96671.0000umn C) and observe what happens to the Gains Chart.%Hit at Random0.0000 0.1000 0.2000 0.3000 0.4000 0.5000 0.6000 0.7000 0.8000 0.900.00000.20000.40000.60000.80001.00001.2000Gains Chart to Evaluate Model Performance%Hit%Hit at Random% Exposed% Hit
000 1.0000
IndividualPurchaseAvg AgeSUMMARY OUTPUT1132151.004214110.397Regression Sta30682-0.165Multiple R405560.237R Square5049180.719Adjusted R Square60651-0.128Standard Error705050.317Observations8128110.974913240.682ANOVA10147120.58711149140.602Regression1212960.806Residual1305970.178Total14063110.2081505820.054160611-0.041Intercept170751-0.347Avg Age1803630.565Distance to Brick1913390.8062005420.1422104950.339SUMMARY OUTPUT220718-0.055230666-0.004Regression Sta2414590.543Multiple R25143110.646R Square2612960.806Adjusted R Square2704830.302Standard Error2805650.186Observations2903110.61630127121.025ANOVARegressionResidualTotalInterceptAvg AgeDistance to BrickDistance to BrickScore (=predicted Y)
Tatistics0.7530.5660.5340.34030dfSSMSFSignificance F24.0792.03917.6390.000273.1210.116297.2CoefficientsStandard Errot StatP-valueLower 95%Upper 95%Lower 95.0%Upper 95.0%1.2650.2734.6430.0000.7061.8250.7061.825-0.0220.005-4.7190.000-0.031-0.012-0.031-0.0120.0290.0142.0960.0460.0010.0580.0010.058Tatistics0.752634760.566459090.534344940.3400160430dfSSMSFSignificance F2 4.0785054 2.03925271 17.6389296 1.2587E-0527 3.1214946 0.11561091297.2CoefficientsStandard Errot StatP-valueLower 95%Upper 95%Lower 95.0%Upper 95.0%1.26536565 0.27252564.6431068 7.9476E-05 0.70618923 1.82454207 0.70618923 1.82454207-0.0218936 0.0046392 -4.7192554 6.4772E-05 -0.0314124 -0.0123747 -0.0314124 -0.01237470.02925298 0.0139534 2.09647731 0.04554475 0.00062297 0.05788298 0.00062297 0.05788298