#Loading Data

library(tidyverse)
## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr     1.2.1     ✔ readr     2.2.0
## ✔ forcats   1.0.1     ✔ stringr   1.6.0
## ✔ ggplot2   4.0.3     ✔ tibble    3.3.1
## ✔ lubridate 1.9.5     ✔ tidyr     1.3.2
## ✔ purrr     1.2.2     
## ── 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 errors
library(MASS)
## 
## Attaching package: 'MASS'
## 
## The following object is masked from 'package:dplyr':
## 
##     select
library(lmtest)
## Loading required package: zoo
## 
## Attaching package: 'zoo'
## 
## The following objects are masked from 'package:base':
## 
##     as.Date, as.Date.numeric
library(car)
## Loading required package: carData
## 
## Attaching package: 'car'
## 
## The following object is masked from 'package:dplyr':
## 
##     recode
## 
## The following object is masked from 'package:purrr':
## 
##     some
hotline_data <- read_csv("hotline_data.csv")
## Rows: 2184 Columns: 11
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (5): State, Timeframe, Call Average Time to Answer, Text Average Time to...
## dbl (6): Number of Routed Calls, Call Answer Rate, Number of Routed Texts, T...
## 
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
#Creating Texas Subset 

texas_data <- hotline_data %>% filter(State == "TX")
#Creating Time Variable

texas_data <- texas_data %>% mutate(Time = row_number())
#Running Basic Descriptive Statistics: Summary

summary(texas_data[, c("Call Answer Rate", "Number of Routed Calls", "Time")])
##  Call Answer Rate Number of Routed Calls      Time     
##  Min.   :0.5617   Min.   :13787          Min.   : 1.0  
##  1st Qu.:0.7623   1st Qu.:14886          1st Qu.:10.5  
##  Median :0.8348   Median :16022          Median :20.0  
##  Mean   :0.7918   Mean   :17377          Mean   :20.0  
##  3rd Qu.:0.8526   3rd Qu.:19294          3rd Qu.:29.5  
##  Max.   :0.8709   Max.   :24509          Max.   :39.0
#Running Basic Descriptive Statistics:Answer Rate Over Time

plot(texas_data$Time,
     texas_data$`Call Answer Rate`,
     xlab = "Time",
     ylab = "Call Answer Rate",
     main = "Texas 988 Call Answer Rate Over Time")

#Running Basic Descriptive Statistics: Answer Rate vs. Call Volume

plot(texas_data$`Number of Routed Calls`,
     texas_data$`Call Answer Rate`,
     xlab = "Number of Routed Calls",
     ylab = "Call Answer Rate",
     main = "Texas 988 Call Volume and Answer Rate")

#Creating a Linear Model

texas_model <- lm(`Call Answer Rate`~`Number of Routed Calls`+ Time,data=texas_data)

summary(texas_model)
## 
## Call:
## lm(formula = `Call Answer Rate` ~ `Number of Routed Calls` + 
##     Time, data = texas_data)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.09730 -0.02269  0.00756  0.02446  0.09029 
## 
## Coefficients:
##                            Estimate Std. Error t value Pr(>|t|)    
## (Intercept)               1.127e+00  7.259e-02  15.528  < 2e-16 ***
## `Number of Routed Calls` -3.457e-05  5.629e-06  -6.142 4.50e-07 ***
## Time                      1.326e-02  1.514e-03   8.763 1.87e-10 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.04839 on 36 degrees of freedom
## Multiple R-squared:  0.7147, Adjusted R-squared:  0.6988 
## F-statistic: 45.08 on 2 and 36 DF,  p-value: 1.571e-10
#Testing Linearity

plot(texas_model, which=1)

Based on visual inspection, linearity is questionable and appears to be violated.

#Testing Independence of Errors

dwtest(texas_model)
## 
##  Durbin-Watson test
## 
## data:  texas_model
## DW = 0.62932, p-value = 3.724e-08
## alternative hypothesis: true autocorrelation is greater than 0

The model’s errors are not independent per the Durbin Watson Test (p-value < 0.05) and appears to be violated.

#Testing Homoscedasticity

plot(texas_model,which=3)

bptest(texas_model)
## 
##  studentized Breusch-Pagan test
## 
## data:  texas_model
## BP = 0.45293, df = 2, p-value = 0.7973

The model is homoscedastistic per the BP-test (p-value > 0.05) and is not violated.

#Testing Normality of Residuals

plot(texas_model,which = 2)

Model appears to be normally distributed based on visual inspection. Some deviations from the reference line are present in the tails.

#Testing Multicolinarity

vif(texas_model)
## `Number of Routed Calls`                     Time 
##                 4.834381                 4.834381

Both independent variables have no severe multicollinearity (VIF < 5).

#Creating a Mitigation Log

texas_model_log<-lm(log(`Call Answer Rate`)~log(`Number of Routed Calls`)+log(Time),data=texas_data)
#Mitigating Linearity

plot(texas_model_log,which = 1)

Model was log-transformed to help with linearity. Based on visual inspection, linearity appears to still not have been met. The assumption may still be violated.

# Copy the data and replace spaces in column names with periods
texas_clean <- texas_data
names(texas_clean) <- make.names(names(texas_clean), unique = TRUE)

# Fit the model using the cleaned names
texas_model_changed <- MASS::rlm(
Call.Answer.Rate ~ Number.of.Routed.Calls + Time,
data = texas_clean
)

summary(texas_model_changed)
## 
## Call: rlm(formula = Call.Answer.Rate ~ Number.of.Routed.Calls + Time, 
##     data = texas_clean)
## Residuals:
##       Min        1Q    Median        3Q       Max 
## -0.099192 -0.022518  0.001704  0.021367  0.095144 
## 
## Coefficients:
##                        Value   Std. Error t value
## (Intercept)             1.1614  0.0786    14.7734
## Number.of.Routed.Calls  0.0000  0.0000    -6.0205
## Time                    0.0135  0.0016     8.2116
## 
## Residual standard error: 0.03351 on 36 degrees of freedom
#Mitigation for Independence of Errors

dwtest(texas_model_changed)
## 
##  Durbin-Watson test
## 
## data:  texas_model_changed
## DW = 0.62932, p-value = 3.724e-08
## alternative hypothesis: true autocorrelation is greater than 0

The Durbin-Watson result did not show that the independence issue disappeared.