Week 10 | Data Dive — GLMs

# Load required libraries
library(tidyverse)

## Warning: package 'ggplot2' was built under R version 4.3.3

## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr     1.1.4     ✔ readr     2.1.4
## ✔ forcats   1.0.0     ✔ stringr   1.5.1
## ✔ ggplot2   3.5.0     ✔ tibble    3.2.1
## ✔ lubridate 1.9.3     ✔ tidyr     1.3.0
## ✔ purrr     1.0.2     
## ── 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(broom)
library(dplyr)

spotify_data <- read.csv("spotify-2023.csv")

n_distinct(spotify_data$mode)

## [1] 2

#Recode 'mode' variable into binary format
spotify_data$mode_binary <- ifelse(spotify_data$mode == "Major", 0, 1)

unique(spotify_data$mode)

## [1] "Major" "Minor"

table(spotify_data$mode_binary)

## 
##   0   1 
## 550 403

# Build a logistic regression model for 'mode

# Select explanatory variables
explanatory_vars <- c("streams", "valence_.", "danceability_.", "energy_.")

# Build logistic regression model
logit_model <- glm(mode_binary ~ streams + valence_. + danceability_. + energy_., 
                   data = spotify_data, 
                   family = binomial)

# Summary of the model
summary(logit_model)

## 
## Call:
## glm(formula = mode_binary ~ streams + valence_. + danceability_. + 
##     energy_., family = binomial, data = spotify_data)
## 
## Coefficients:
##                  Estimate Std. Error z value Pr(>|z|)    
## (Intercept)    -1.744e+00  3.986e-01  -4.377 1.21e-05 ***
## streams        -1.016e-10  1.194e-10  -0.851 0.395031    
## valence_.      -4.412e-06  3.242e-03  -0.001 0.998914    
## danceability_.  1.851e-02  5.089e-03   3.638 0.000275 ***
## energy_.        3.700e-03  4.344e-03   0.852 0.394297    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for binomial family taken to be 1)
## 
##     Null deviance: 1298.4  on 952  degrees of freedom
## Residual deviance: 1278.3  on 948  degrees of freedom
## AIC: 1288.3
## 
## Number of Fisher Scoring iterations: 4

• Streams: This coefficient is very small and statistically insignificant, meaning the number of streams has little impact on predicting the music genre

• Valence_ and Energy_: These coefficients are also statistically insignificant, suggesting these features don’t strongly influence the model’s prediction.

• Danceability_: This coefficient is positive and statistically significant. A higher danceability score increases the predicted probability of the song belonging to a dance-oriented genre.

# Calculate confidence interval for 'danceability_' coefficient
ci_danceability <- confint(logit_model, "danceability_.")

## Waiting for profiling to be done...

ci_danceability

##       2.5 %      97.5 % 
## 0.008605308 0.028572961

The confidence interval for the coefficient of the ‘danceability_’ variable is as follows:

• Lower bound (2.5%): 0.008605308

• Upper bound (97.5%): 0.028572961

This indicates that we are 95% confident that the true coefficient of ‘danceability_’ lies within the range of approximately 0.0086 to 0.0286. In other words, for each unit increase in danceability percentage, the change in the log odds of a song being in ‘Minor’ mode is expected to lie within this interval.