Analysis of IMDB Rating with GLM - Logistic Regression

In this analysis, the research question is to investigate which properties of films influence whether IMDB rating exceeding 7 or not. The Generalized Linear Model for binary response variables, Logistic Regression, will be used to investigate the relationship between binary rating (i.e., 1 if greater than 7 and 0 otherwise) and film properties.

Installation and Setup

R and quarto installation is required as the main tools for doing the statistical analysis and reporting. Here is the list of R packages which are needed to be installed in order to run the main .qmd file in the local machine:

install.packages(tidyverse)
install.packages(gt)
install.packages(skimr)
install.packages(knitr)
install.packages(corrplot)
install.packages(ggplot2)
install.packages(gridExtra)
install.packages(dplyr)
install.packages(stats)
install.packages(jtools)
install.packages(sjPlot)
install.packages(broom)
install.packages(huxtable)
install.packages(lmtest)
install.packages(zoo)

Data

IMDB rating dataset which is preprocessed by Data Analytics (Statistics) department from University of Glasgow as the playground dataset for learning purposes.

Film ID	Year	Length	Budget	Votes	Genre	Rating
49834	1963	107	11.4	225	Romance	3.1

Description:

Film ID: The unique identifier for the film.
Year: Year of release of the film in cinemas.
Length: Duration (in minutes).
Budget: Budget for the films production (in $1000000).
Votes: Number of positive votes received by viewers.
Genre: Genre of the film.
Rating: IMDB rating from 0-10.

Code structure

/DAS-Group-10
├── README.md
├── data/
│   └── dataset10.csv
├── pdf/
│   ├── Group_10_Presentation.pdf
│   ├── Group_10_qmd.pdf
│   └── raw/Group_10_Presentation.pptx
└── quarto/
    ├── R
    └── Group_10_Analysis.qmd

Results

In summary, budget and length have significant relationship to binary rating, whether greater than 7 or not. For Genre, two of them are not statistically significant: Animation and Romance.

Model selection steps

Hypothesis testing on the estimated coefficients using P-value and CIs.
Model comparison using AIC and BIC as the criteria.
Goodness of fit using Hosmer-Lemeshow test.

Future Work

Future studies should consider interaction terms between movie length and genre which provide insights into the nuanced effects these variables have on ratings.