logistic regression
ESHA SOHAIL
2024-03-20
Explore the data to find any Outliers / suspected outliers using modified box and whisker plot
Compare the models and prepare a report (i) splitting 20% with 30% (ii) 5-fold with 10-fold cross validation #### import library
library(ggplot2)## Warning: package 'ggplot2' was built under R version 4.3.3library(tidyverse)## Warning: package 'tidyverse' was built under R version 4.3.3## Warning: package 'tidyr' was built under R version 4.3.3## Warning: package 'purrr' was built under R version 4.3.3## Warning: package 'dplyr' was built under R version 4.3.3## Warning: package 'forcats' was built under R version 4.3.3## Warning: package 'lubridate' was built under R version 4.3.3## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr 1.1.4 ✔ readr 2.1.5
## ✔ forcats 1.0.0 ✔ stringr 1.5.1
## ✔ lubridate 1.9.3 ✔ tibble 3.2.1
## ✔ purrr 1.0.2 ✔ tidyr 1.3.1
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errorslibrary(dplyr)
library(caTools)## Warning: package 'caTools' was built under R version 4.3.3library(caret)## Warning: package 'caret' was built under R version 4.3.3## Loading required package: lattice
##
## Attaching package: 'caret'
##
## The following object is masked from 'package:purrr':
##
## liftdonors<-read.csv("C:/Users/PMLS/Downloads/donors.csv")
glimpse(donors)## Rows: 95,412
## Columns: 22
## $ age <int> 60, 46, NA, 70, 78, NA, 38, NA, NA, 65, NA, 75…
## $ numberChildren <int> NA, 1, NA, NA, 1, NA, 1, NA, NA, NA, NA, NA, 2…
## $ incomeRating <int> NA, 6, 3, 1, 3, NA, 4, 2, 3, NA, 2, 1, 4, NA, …
## $ wealthRating <int> NA, 9, 1, 4, 2, NA, 6, 9, 2, NA, 0, 5, 2, NA, …
## $ mailOrderPurchases <int> 0, 16, 2, 2, 60, 0, 0, 1, 0, 0, 0, 3, 16, 0, 1…
## $ totalGivingAmount <dbl> 240, 47, 202, 109, 254, 51, 107, 31, 199, 28, …
## $ numberGifts <int> 31, 3, 27, 16, 37, 4, 14, 5, 11, 3, 1, 2, 9, 1…
## $ smallestGiftAmount <dbl> 5, 10, 2, 2, 3, 10, 3, 5, 10, 3, 20, 10, 4, 5,…
## $ largestGiftAmount <dbl> 12, 25, 16, 11, 15, 16, 12, 11, 22, 15, 20, 15…
## $ averageGiftAmount <dbl> 7.741935, 15.666667, 7.481481, 6.812500, 6.864…
## $ yearsSinceFirstDonation <int> 8, 3, 7, 10, 11, 3, 10, 3, 9, 3, 1, 1, 8, 5, 4…
## $ monthsSinceLastDonation <int> 14, 14, 14, 14, 13, 20, 22, 18, 19, 22, 12, 14…
## $ inHouseDonor <lgl> FALSE, FALSE, FALSE, FALSE, TRUE, FALSE, FALSE…
## $ plannedGivingDonor <lgl> FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALS…
## $ sweepstakesDonor <lgl> FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALS…
## $ P3Donor <lgl> FALSE, FALSE, FALSE, FALSE, TRUE, FALSE, FALSE…
## $ state <chr> "IL", "CA", "NC", "CA", "FL", "AL", "IN", "LA"…
## $ urbanicity <chr> "town", "suburb", "rural", "rural", "suburb", …
## $ socioEconomicStatus <chr> "average", "highest", "average", "average", "a…
## $ isHomeowner <lgl> NA, TRUE, NA, NA, TRUE, NA, TRUE, NA, NA, NA, …
## $ gender <chr> "female", "male", "male", "female", "female", …
## $ respondedMailing <lgl> FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALS…Dealing with missing values
donors <- donors %>%
mutate(incomeRating = as.character(incomeRating)) %>%
mutate(incomeRating = as.factor(ifelse(is.na(incomeRating), 'UNK',
incomeRating)))
donors <- donors %>%
mutate(wealthRating = as.character(wealthRating)) %>%
mutate(wealthRating = as.factor(ifelse(is.na(wealthRating), 'UNK',
wealthRating))) %>%
mutate(urbanicity = as.character(urbanicity)) %>%
mutate(urbanicity = as.factor(ifelse(is.na(urbanicity), 'UNK',
urbanicity))) %>%
mutate(socioEconomicStatus = as.character(socioEconomicStatus)) %>%
mutate(socioEconomicStatus = as.factor(ifelse(is.na(socioEconomicStatus), 'UNK', socioEconomicStatus))) %>%
mutate(isHomeowner = as.character(isHomeowner)) %>%
mutate(isHomeowner = as.factor(ifelse(is.na(isHomeowner), 'UNK',
isHomeowner))) %>%
mutate(gender = as.character(gender)) %>%
mutate(gender = as.factor(ifelse(is.na(gender), 'UNK', gender)))
glimpse(donors)## Rows: 95,412
## Columns: 22
## $ age <int> 60, 46, NA, 70, 78, NA, 38, NA, NA, 65, NA, 75…
## $ numberChildren <int> NA, 1, NA, NA, 1, NA, 1, NA, NA, NA, NA, NA, 2…
## $ incomeRating <fct> UNK, 6, 3, 1, 3, UNK, 4, 2, 3, UNK, 2, 1, 4, U…
## $ wealthRating <fct> UNK, 9, 1, 4, 2, UNK, 6, 9, 2, UNK, 0, 5, 2, U…
## $ mailOrderPurchases <int> 0, 16, 2, 2, 60, 0, 0, 1, 0, 0, 0, 3, 16, 0, 1…
## $ totalGivingAmount <dbl> 240, 47, 202, 109, 254, 51, 107, 31, 199, 28, …
## $ numberGifts <int> 31, 3, 27, 16, 37, 4, 14, 5, 11, 3, 1, 2, 9, 1…
## $ smallestGiftAmount <dbl> 5, 10, 2, 2, 3, 10, 3, 5, 10, 3, 20, 10, 4, 5,…
## $ largestGiftAmount <dbl> 12, 25, 16, 11, 15, 16, 12, 11, 22, 15, 20, 15…
## $ averageGiftAmount <dbl> 7.741935, 15.666667, 7.481481, 6.812500, 6.864…
## $ yearsSinceFirstDonation <int> 8, 3, 7, 10, 11, 3, 10, 3, 9, 3, 1, 1, 8, 5, 4…
## $ monthsSinceLastDonation <int> 14, 14, 14, 14, 13, 20, 22, 18, 19, 22, 12, 14…
## $ inHouseDonor <lgl> FALSE, FALSE, FALSE, FALSE, TRUE, FALSE, FALSE…
## $ plannedGivingDonor <lgl> FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALS…
## $ sweepstakesDonor <lgl> FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALS…
## $ P3Donor <lgl> FALSE, FALSE, FALSE, FALSE, TRUE, FALSE, FALSE…
## $ state <chr> "IL", "CA", "NC", "CA", "FL", "AL", "IN", "LA"…
## $ urbanicity <fct> town, suburb, rural, rural, suburb, town, town…
## $ socioEconomicStatus <fct> average, highest, average, average, average, a…
## $ isHomeowner <fct> UNK, TRUE, UNK, UNK, TRUE, UNK, TRUE, UNK, UNK…
## $ gender <fct> female, male, male, female, female, UNK, femal…
## $ respondedMailing <lgl> FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALS…continous_numeric_vars <- c("age")
for (var in continous_numeric_vars) {
donors[[var]][is.na(donors[[var]])] <-mean(donors[[var]], na.rm = TRUE)
}
numeric_vars <- c("numberChildren")
for (var in numeric_vars) {
donors[[var]][is.na(donors[[var]])] <-median(donors[[var]], na.rm = TRUE)
}glimpse(donors)## Rows: 95,412
## Columns: 22
## $ age <dbl> 60.00000, 46.00000, 61.61165, 70.00000, 78.000…
## $ numberChildren <dbl> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1…
## $ incomeRating <fct> UNK, 6, 3, 1, 3, UNK, 4, 2, 3, UNK, 2, 1, 4, U…
## $ wealthRating <fct> UNK, 9, 1, 4, 2, UNK, 6, 9, 2, UNK, 0, 5, 2, U…
## $ mailOrderPurchases <int> 0, 16, 2, 2, 60, 0, 0, 1, 0, 0, 0, 3, 16, 0, 1…
## $ totalGivingAmount <dbl> 240, 47, 202, 109, 254, 51, 107, 31, 199, 28, …
## $ numberGifts <int> 31, 3, 27, 16, 37, 4, 14, 5, 11, 3, 1, 2, 9, 1…
## $ smallestGiftAmount <dbl> 5, 10, 2, 2, 3, 10, 3, 5, 10, 3, 20, 10, 4, 5,…
## $ largestGiftAmount <dbl> 12, 25, 16, 11, 15, 16, 12, 11, 22, 15, 20, 15…
## $ averageGiftAmount <dbl> 7.741935, 15.666667, 7.481481, 6.812500, 6.864…
## $ yearsSinceFirstDonation <int> 8, 3, 7, 10, 11, 3, 10, 3, 9, 3, 1, 1, 8, 5, 4…
## $ monthsSinceLastDonation <int> 14, 14, 14, 14, 13, 20, 22, 18, 19, 22, 12, 14…
## $ inHouseDonor <lgl> FALSE, FALSE, FALSE, FALSE, TRUE, FALSE, FALSE…
## $ plannedGivingDonor <lgl> FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALS…
## $ sweepstakesDonor <lgl> FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALS…
## $ P3Donor <lgl> FALSE, FALSE, FALSE, FALSE, TRUE, FALSE, FALSE…
## $ state <chr> "IL", "CA", "NC", "CA", "FL", "AL", "IN", "LA"…
## $ urbanicity <fct> town, suburb, rural, rural, suburb, town, town…
## $ socioEconomicStatus <fct> average, highest, average, average, average, a…
## $ isHomeowner <fct> UNK, TRUE, UNK, UNK, TRUE, UNK, TRUE, UNK, UNK…
## $ gender <fct> female, male, male, female, female, UNK, femal…
## $ respondedMailing <lgl> FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALS…Making boxplots of continuous variables for outlier detection
variables_of_interest <- c("age", "numberChildren", "mailOrderPurchases","totalGivingAmount")
par(mfrow = c(2,2))
for (variable in variables_of_interest) {
boxplot(donors[[variable]], main = variable, ylab = variable, col = "pink")
}
variables_of_interest<-c("numberGifts","smallestGiftAmount","largestGiftAmount","averageGiftAmount")
par(mfrow = c(2,2))
for (variable in variables_of_interest) {
boxplot(donors[[variable]], main = variable, ylab = variable, col = "yellow")
}
variables_of_interest <- c("yearsSinceFirstDonation","monthsSinceLastDonation")
par(mfrow = c(1,1))
for (variable in variables_of_interest) {
boxplot(donors[[variable]], main = variable, ylab = variable, col = "brown")
}
#### Making whiskers plot for outlier detection
variables_of_interest <- c("age", "numberChildren", "mailOrderPurchases","totalGivingAmount","numberGifts","smallestGiftAmount","largestGiftAmount","averageGiftAmount","yearsSinceFirstDonation","monthsSinceLastDonation")
par(mfrow = c(2, 5), mar = c(5, 4, 2, 2))
for (variable in variables_of_interest) {
boxplot(donors[[variable]] ~ donors$respondedMailing, data = donors,
main = paste("Whisker Plot of", variable, "by Target"),
xlab = "Target", ylab = variable, col = "brown")
}
#### Now we will remove outliers
donors <- donors %>%
mutate(max1 = quantile(mailOrderPurchases, .75) + (1.5 *IQR(mailOrderPurchases))) %>%
mutate(max2 = quantile(totalGivingAmount, .75) + (1.5 *
IQR(totalGivingAmount))) %>%
mutate(max3 = quantile(numberGifts, .75) + (1.5 * IQR(numberGifts))) %>%
mutate(max4 = quantile(smallestGiftAmount, .75) + (1.5 *
IQR(smallestGiftAmount))) %>%
mutate(max5 = quantile(largestGiftAmount, .75) + (1.5 *
IQR(largestGiftAmount))) %>%
mutate(max6 = quantile(averageGiftAmount, .75) + (1.5 *
IQR(averageGiftAmount))) %>%
filter(mailOrderPurchases <= max1) %>%
filter(totalGivingAmount <= max2) %>%
filter(numberGifts <= max3) %>%
filter(smallestGiftAmount <= max4) %>%
filter(largestGiftAmount <= max5) %>%
filter(averageGiftAmount <= max6) %>%
select(-max1,-max2,-max3,-max4,-max5,-max6)Making a function to perform logistics regression and finding an Accuracy score
set.seed(1234)
donors$respondedMailing <- factor(donors$respondedMailing)
evaluate_logistic_regression <- function(train_data, test_data) {
model <- glm(respondedMailing ~ ., data = train_data, family = "binomial")
summary(model)
}20 and 80 split
split_20_80 <- sample.split(donors$respondedMailing, SplitRatio = 0.8)
train_data_20_80 <- donors[split_20_80, ]
test_data_20_80 <- donors[!split_20_80, ]
summary_20_80 <- evaluate_logistic_regression(train_data_20_80, test_data_20_80)
print(summary_20_80)##
## Call:
## glm(formula = respondedMailing ~ ., family = "binomial", data = train_data)
##
## Coefficients: (1 not defined because of singularities)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 4.052e-01 7.369e-01 0.550 0.582412
## age -1.055e-03 1.493e-03 -0.707 0.479837
## numberChildren -1.182e-01 6.385e-02 -1.850 0.064250 .
## incomeRating2 9.399e-02 9.105e-02 1.032 0.301932
## incomeRating3 1.857e-01 9.959e-02 1.865 0.062219 .
## incomeRating4 2.344e-01 9.424e-02 2.487 0.012885 *
## incomeRating5 2.459e-01 9.163e-02 2.684 0.007269 **
## incomeRating6 2.258e-01 1.078e-01 2.095 0.036205 *
## incomeRating7 3.510e-01 1.080e-01 3.252 0.001148 **
## incomeRatingUNK 3.176e-01 9.011e-02 3.524 0.000425 ***
## wealthRating1 -1.520e-01 1.797e-01 -0.846 0.397717
## wealthRating2 2.113e-02 1.697e-01 0.124 0.900927
## wealthRating3 1.334e-02 1.688e-01 0.079 0.936975
## wealthRating4 -1.619e-01 1.710e-01 -0.946 0.343964
## wealthRating5 -1.667e-01 1.689e-01 -0.987 0.323572
## wealthRating6 -1.102e-01 1.664e-01 -0.662 0.507954
## wealthRating7 -1.501e-01 1.668e-01 -0.900 0.368314
## wealthRating8 -1.411e-01 1.663e-01 -0.848 0.396346
## wealthRating9 -6.277e-02 1.675e-01 -0.375 0.707777
## wealthRatingUNK -1.382e-01 1.487e-01 -0.929 0.352713
## mailOrderPurchases -4.370e-03 1.387e-02 -0.315 0.752779
## totalGivingAmount 1.471e-03 9.497e-04 1.549 0.121399
## numberGifts 4.566e-03 8.979e-03 0.508 0.611111
## smallestGiftAmount -7.559e-03 1.006e-02 -0.751 0.452429
## largestGiftAmount -4.611e-02 7.326e-03 -6.295 3.08e-10 ***
## averageGiftAmount 4.848e-03 1.639e-02 0.296 0.767444
## yearsSinceFirstDonation 5.743e-03 9.811e-03 0.585 0.558353
## monthsSinceLastDonation -3.973e-02 5.755e-03 -6.904 5.07e-12 ***
## inHouseDonorTRUE 5.934e-02 8.762e-02 0.677 0.498244
## plannedGivingDonorTRUE 3.563e-01 7.463e-01 0.477 0.633018
## sweepstakesDonorTRUE -5.519e-01 1.730e-01 -3.191 0.001418 **
## P3DonorTRUE 1.062e-01 1.316e-01 0.807 0.419640
## stateAE -1.362e+01 2.928e+02 -0.047 0.962889
## stateAK -3.426e+00 9.936e-01 -3.448 0.000564 ***
## stateAL -2.069e+00 7.065e-01 -2.929 0.003404 **
## stateAP -1.851e+00 9.093e-01 -2.035 0.041817 *
## stateAR -2.100e+00 7.179e-01 -2.925 0.003443 **
## stateAZ -1.843e+00 7.008e-01 -2.630 0.008533 **
## stateCA -1.875e+00 6.925e-01 -2.707 0.006795 **
## stateCO -2.114e+00 7.062e-01 -2.993 0.002760 **
## stateCT -1.987e+00 1.244e+00 -1.597 0.110320
## stateDC -1.321e+01 8.827e+02 -0.015 0.988063
## stateDE -1.359e+01 8.827e+02 -0.015 0.987719
## stateFL -2.043e+00 6.948e-01 -2.941 0.003270 **
## stateGA -2.465e+00 7.029e-01 -3.508 0.000452 ***
## stateGU -1.341e-01 1.416e+00 -0.095 0.924541
## stateHI -1.559e+00 7.307e-01 -2.134 0.032839 *
## stateIA -2.243e+00 7.160e-01 -3.133 0.001728 **
## stateID -2.186e+00 7.452e-01 -2.934 0.003351 **
## stateIL -2.140e+00 6.960e-01 -3.075 0.002108 **
## stateIN -2.252e+00 7.023e-01 -3.206 0.001347 **
## stateKS -2.127e+00 7.127e-01 -2.984 0.002842 **
## stateKY -2.033e+00 7.072e-01 -2.875 0.004046 **
## stateLA -2.164e+00 7.088e-01 -3.053 0.002266 **
## stateMA -1.358e+01 2.066e+02 -0.066 0.947581
## stateMD -8.037e-01 9.412e-01 -0.854 0.393159
## stateME -1.367e+01 3.592e+02 -0.038 0.969636
## stateMI -1.998e+00 6.962e-01 -2.870 0.004107 **
## stateMN -2.424e+00 7.075e-01 -3.426 0.000613 ***
## stateMO -2.185e+00 7.024e-01 -3.110 0.001869 **
## stateMS -2.226e+00 7.222e-01 -3.083 0.002050 **
## stateMT -2.034e+00 7.367e-01 -2.761 0.005770 **
## stateNC -2.085e+00 6.988e-01 -2.984 0.002847 **
## stateND -2.010e+00 7.827e-01 -2.568 0.010217 *
## stateNE -2.203e+00 7.292e-01 -3.021 0.002519 **
## stateNH -1.384e+01 5.089e+02 -0.027 0.978300
## stateNJ -1.396e+00 1.028e+00 -1.357 0.174625
## stateNM -1.972e+00 7.190e-01 -2.743 0.006079 **
## stateNV -1.984e+00 7.147e-01 -2.777 0.005493 **
## stateNY -3.075e+00 1.225e+00 -2.509 0.012103 *
## stateOH -1.361e+01 1.438e+02 -0.095 0.924568
## stateOK -2.471e+00 7.153e-01 -3.454 0.000552 ***
## stateOR -1.715e+00 7.012e-01 -2.446 0.014443 *
## statePA -2.340e+00 1.234e+00 -1.896 0.057969 .
## stateRI -1.340e+01 3.572e+02 -0.038 0.970079
## stateSC -2.044e+00 7.063e-01 -2.894 0.003805 **
## stateSD -1.545e+00 7.408e-01 -2.086 0.036958 *
## stateTN -2.387e+00 7.056e-01 -3.383 0.000717 ***
## stateTX -2.149e+00 6.955e-01 -3.090 0.002001 **
## stateUT -2.342e+00 7.476e-01 -3.133 0.001729 **
## stateVA -1.359e+01 1.479e+02 -0.092 0.926743
## stateVI -1.362e+01 3.925e+02 -0.035 0.972309
## stateVT -1.364e+01 4.378e+02 -0.031 0.975153
## stateWA -2.048e+00 6.992e-01 -2.929 0.003400 **
## stateWI -2.033e+00 7.012e-01 -2.899 0.003743 **
## stateWV 1.013e-02 1.416e+00 0.007 0.994292
## stateWY -2.224e+00 7.938e-01 -2.802 0.005086 **
## urbanicityrural -8.005e-02 6.530e-02 -1.226 0.220266
## urbanicitysuburb 2.927e-02 6.086e-02 0.481 0.630506
## urbanicitytown -2.033e-02 6.293e-02 -0.323 0.746625
## urbanicityUNK -3.328e-02 1.357e-01 -0.245 0.806240
## urbanicityurban -2.180e-01 7.444e-02 -2.928 0.003409 **
## socioEconomicStatushighest 9.219e-02 5.326e-02 1.731 0.083458 .
## socioEconomicStatuslowest -1.392e-01 6.148e-02 -2.265 0.023524 *
## socioEconomicStatusUNK NA NA NA NA
## isHomeownerUNK -6.260e-02 5.458e-02 -1.147 0.251414
## genderjoint 4.262e-01 2.648e-01 1.609 0.107554
## gendermale -1.610e-02 4.094e-02 -0.393 0.694238
## genderUNK -2.787e-02 8.972e-02 -0.311 0.756063
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 22317 on 56007 degrees of freedom
## Residual deviance: 21830 on 55910 degrees of freedom
## AIC: 22026
##
## Number of Fisher Scoring iterations: 1330 and 70 split
split_30_70 <- sample.split(donors$respondedMailing, SplitRatio = 0.7)
train_data_30_70 <- donors[split_30_70, ]
test_data_30_70 <- donors[!split_30_70, ]
summary_30_70 <- evaluate_logistic_regression(train_data_30_70, test_data_30_70)
print(summary_30_70)##
## Call:
## glm(formula = respondedMailing ~ ., family = "binomial", data = train_data)
##
## Coefficients: (1 not defined because of singularities)
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 3.017e-01 8.973e-01 0.336 0.73667
## age -2.618e-03 1.592e-03 -1.645 0.10003
## numberChildren -1.661e-01 6.965e-02 -2.385 0.01706 *
## incomeRating2 1.150e-01 9.813e-02 1.172 0.24114
## incomeRating3 2.107e-01 1.069e-01 1.970 0.04881 *
## incomeRating4 2.075e-01 1.020e-01 2.034 0.04194 *
## incomeRating5 2.765e-01 9.840e-02 2.810 0.00496 **
## incomeRating6 2.495e-01 1.150e-01 2.171 0.02996 *
## incomeRating7 4.924e-01 1.135e-01 4.338 1.44e-05 ***
## incomeRatingUNK 2.294e-01 9.669e-02 2.372 0.01769 *
## wealthRating1 1.058e-02 2.047e-01 0.052 0.95878
## wealthRating2 4.545e-02 1.964e-01 0.231 0.81695
## wealthRating3 1.303e-01 1.928e-01 0.676 0.49933
## wealthRating4 -1.845e-02 1.942e-01 -0.095 0.92430
## wealthRating5 1.068e-01 1.886e-01 0.566 0.57119
## wealthRating6 2.103e-01 1.863e-01 1.129 0.25907
## wealthRating7 7.495e-03 1.890e-01 0.040 0.96837
## wealthRating8 1.030e-01 1.865e-01 0.552 0.58063
## wealthRating9 5.192e-02 1.894e-01 0.274 0.78402
## wealthRatingUNK 9.309e-02 1.704e-01 0.546 0.58483
## mailOrderPurchases -1.602e-03 1.491e-02 -0.107 0.91444
## totalGivingAmount 2.059e-03 1.023e-03 2.013 0.04411 *
## numberGifts 1.734e-03 9.683e-03 0.179 0.85791
## smallestGiftAmount -1.441e-02 1.082e-02 -1.331 0.18304
## largestGiftAmount -4.674e-02 7.791e-03 -5.999 1.98e-09 ***
## averageGiftAmount -6.862e-04 1.758e-02 -0.039 0.96886
## yearsSinceFirstDonation 5.705e-04 1.045e-02 0.055 0.95647
## monthsSinceLastDonation -3.308e-02 6.112e-03 -5.412 6.22e-08 ***
## inHouseDonorTRUE 9.393e-02 9.274e-02 1.013 0.31117
## plannedGivingDonorTRUE 5.117e-01 7.490e-01 0.683 0.49453
## sweepstakesDonorTRUE -4.377e-01 1.806e-01 -2.423 0.01539 *
## P3DonorTRUE 6.523e-02 1.425e-01 0.458 0.64714
## stateAE -1.382e+01 3.591e+02 -0.038 0.96930
## stateAK -2.666e+00 9.948e-01 -2.680 0.00736 **
## stateAL -2.062e+00 8.680e-01 -2.375 0.01755 *
## stateAP -2.889e+00 1.322e+00 -2.184 0.02895 *
## stateAR -2.120e+00 8.801e-01 -2.409 0.01599 *
## stateAZ -1.962e+00 8.637e-01 -2.271 0.02313 *
## stateCA -1.871e+00 8.552e-01 -2.187 0.02873 *
## stateCO -2.140e+00 8.679e-01 -2.466 0.01365 *
## stateCT -1.861e+00 1.346e+00 -1.382 0.16686
## stateDC -1.309e+01 8.827e+02 -0.015 0.98817
## stateDE -1.370e+01 6.237e+02 -0.022 0.98248
## stateFL -2.169e+00 8.576e-01 -2.529 0.01144 *
## stateGA -2.381e+00 8.642e-01 -2.755 0.00587 **
## stateGU 2.755e-01 1.656e+00 0.166 0.86789
## stateHI -1.793e+00 8.958e-01 -2.002 0.04533 *
## stateIA -2.304e+00 8.786e-01 -2.622 0.00875 **
## stateID -2.046e+00 8.987e-01 -2.277 0.02281 *
## stateIL -2.271e+00 8.589e-01 -2.644 0.00819 **
## stateIN -2.211e+00 8.640e-01 -2.559 0.01049 *
## stateKS -2.025e+00 8.726e-01 -2.321 0.02031 *
## stateKY -2.203e+00 8.709e-01 -2.530 0.01141 *
## stateLA -2.389e+00 8.734e-01 -2.735 0.00623 **
## stateMA -1.363e+01 2.010e+02 -0.068 0.94592
## stateMD -1.737e+00 1.348e+00 -1.289 0.19743
## stateME -1.360e+01 3.596e+02 -0.038 0.96984
## stateMI -2.085e+00 8.589e-01 -2.427 0.01521 *
## stateMN -2.319e+00 8.679e-01 -2.671 0.00755 **
## stateMO -2.053e+00 8.633e-01 -2.378 0.01740 *
## stateMS -2.111e+00 8.805e-01 -2.397 0.01652 *
## stateMT -1.912e+00 8.964e-01 -2.133 0.03292 *
## stateNC -2.078e+00 8.610e-01 -2.413 0.01582 *
## stateND -1.882e+00 9.307e-01 -2.022 0.04316 *
## stateNE -2.220e+00 8.904e-01 -2.493 0.01268 *
## stateNH -1.395e+01 4.409e+02 -0.032 0.97476
## stateNJ -1.529e+00 1.354e+00 -1.129 0.25892
## stateNM -2.052e+00 8.825e-01 -2.325 0.02006 *
## stateNV -2.138e+00 8.798e-01 -2.429 0.01512 *
## stateNY -2.935e+00 1.325e+00 -2.215 0.02679 *
## stateOH -1.364e+01 1.709e+02 -0.080 0.93638
## stateOK -2.329e+00 8.739e-01 -2.665 0.00770 **
## stateOR -1.795e+00 8.638e-01 -2.078 0.03767 *
## statePA -2.037e+00 1.340e+00 -1.520 0.12843
## stateRI -1.342e+01 3.896e+02 -0.034 0.97252
## stateSC -1.966e+00 8.669e-01 -2.267 0.02337 *
## stateSD -1.508e+00 9.005e-01 -1.675 0.09398 .
## stateTN -2.310e+00 8.662e-01 -2.667 0.00766 **
## stateTX -2.048e+00 8.576e-01 -2.388 0.01696 *
## stateUT -2.136e+00 8.975e-01 -2.380 0.01730 *
## stateVA -2.574e+00 1.329e+00 -1.937 0.05274 .
## stateVI -1.348e+01 5.085e+02 -0.027 0.97885
## stateVT -1.390e+01 6.232e+02 -0.022 0.98221
## stateWA -1.977e+00 8.606e-01 -2.297 0.02164 *
## stateWI -2.033e+00 8.634e-01 -2.355 0.01851 *
## stateWV 9.837e-01 1.662e+00 0.592 0.55406
## stateWY -1.892e+00 9.303e-01 -2.033 0.04202 *
## urbanicityrural -8.421e-02 7.018e-02 -1.200 0.23020
## urbanicitysuburb 4.636e-02 6.527e-02 0.710 0.47752
## urbanicitytown -1.768e-02 6.754e-02 -0.262 0.79354
## urbanicityUNK 1.028e-01 1.398e-01 0.735 0.46207
## urbanicityurban -1.357e-01 7.936e-02 -1.709 0.08737 .
## socioEconomicStatushighest 1.037e-01 5.658e-02 1.834 0.06671 .
## socioEconomicStatuslowest -7.073e-02 6.590e-02 -1.073 0.28318
## socioEconomicStatusUNK NA NA NA NA
## isHomeownerUNK -1.621e-02 5.792e-02 -0.280 0.77954
## genderjoint 1.895e-01 3.154e-01 0.601 0.54798
## gendermale 2.737e-02 4.369e-02 0.627 0.53097
## genderUNK 5.843e-02 9.398e-02 0.622 0.53408
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 19532 on 49006 degrees of freedom
## Residual deviance: 19093 on 48909 degrees of freedom
## AIC: 19289
##
## Number of Fisher Scoring iterations: 135 fold
####{r} set.seed(1234) cv_5_fold <- trainControl(method = "cv", number = 5) summary_5_fold <- train(respondedMailing ~ ., data = donors, method = "glm", trControl = cv_5_fold, family = "binomial")$finalModel print(summary_5_fold)
#### 10 folds
###{r} set.seed(1234) cv_10_fold <- trainControl(method = "cv", number = 10) summary_10_fold <- train(respondedMailing ~ ., data = donors, method = "glm", trControl = cv_10_fold, family = "binomial")$finalModel print(summary_10_fold)
### Findings and Interpretation
A lower null deviance indicates that the model fits the data well even without any predictors.
The null deviance is useful for comparing the fit of the model to a model with no predictors (null model).
Lower residual deviance indicates better fit of the model to the data.
Residual deviance should ideally be significantly lower than the null deviance, indicating that the model improves the prediction compared to the null model.
A lower AIC value indicates a better model, with a good balance between fit and complexity.
Models with smaller AIC values are preferred over models with larger AIC values.
AIC can be used to compare different models, with the model having the lowest AIC being considered the best among the alternatives.