Netflix Data Final Analysis: Duration, Release Year, and Growth Trends

Introduction

           The purpose of this project is to examine patterns in Netflix content using a dataset of 8,807 titles. This analysis focuses on three main questions: whether movies and TV shows are released in different average years, whether movie duration varies across rating categories such as G, PG-13, and TV-MA, and whether the number of titles added to Netflix has increased over time. Together, these questions help explore how Netflix’s content has evolved, how rating categories relate to the length of films, and the overall growth of Netflix’s library in recent years.

          This project uses the same inferential techniques practiced in class, including two-sample t-tests to compare group means, ANOVA to test for differences across multiple categories, and linear regression to examine trends over time. Together, these statistical methods allow us to draw evidence-based conclusions from the data rather than relying solely on descriptive summaries.

Dataset Preparation

         The original dataset (netflix_titles.csv) contains 8,807 entries. To prepare the data for analysis, movie durations were cleaned by extracting the numeric values from the “duration” column, converting entries such as “90 min” into a usable numeric format. The date_added variable was also converted into a proper date format, allowing the year_added field to be extracted for trend analysis. After cleaning, separate datasets were created for movies and for examining year-added patterns to ensure accurate and focused statistical testing.

T-test: Release Year for Movies vs TV shows

          The hypotheses for this part of the analysis examine whether the average release year differs between movies and TV shows. The null hypothesis (H₀) states that movies and TV shows share the same mean release year, while the alternative hypothesis (H₁) proposes that their average release years are different.

A screenshot of a computer code AI-generated content may be incorrect.

Results:

              The t-test results showed a substantial difference in release years between movies and TV shows. The test statistic was t = −20.976 with a highly significant p-value less than 2.2e-16, indicating strong evidence against the null hypothesis. The 95% confidence interval for the difference in means ranged from −3.81 to −3.16, confirming that the difference is both statistically significant and practically meaningful. On average, movies in the dataset were released in 2013.12, while TV shows had a more recent mean release year of 2016.61.

Interpretation:

           The results show a highly significant difference in release year between movies and TV shows (p < 0.001). TV shows on Netflix tend to be released more recently (mean = 2016.6) compared to movies (mean = 2013.1). This suggests Netflix adds newer TV shows more frequently than it adds newer movies.     

AVONA: Movie Duration Across Ratings

         The hypotheses for this part of the analysis focus on whether movie duration varies across different rating categories. The null hypothesis (H₀) states that the mean movie duration is the same for all rating categories, while the alternative hypothesis (H₁) proposes that at least one rating category differs in its average duration.

A screenshot of a computer AI-generated content may be incorrect.

Results:

        The ANOVA results indicated a significant difference in movie duration across the various rating categories. The test produced an F-statistic of F(13, 6112) = 98.11 with a p-value less than 2e-16, providing strong evidence that not all rating categories share the same mean duration. This confirms that movie length varies meaningfully depending on the rating assigned.

A graph with black and white lines and dots AI-generated content may be incorrect.

Interpretation: 

             The ANOVA test indicates a statistically significant difference in movie duration across rating categories (p < 0.001). This means movie length varies depending on the film’s rating. The boxplot supports this by showing categories such as NC-17 and TV-MA tending to have longer films, while TV-Y and TV-G movies are much shorter on average.

Regression: Number of Titles Added per year

             The regression model used to analyze the trend in Netflix content additions over time was specified as: count_of_titles_added predicted by year_added, expressed as count_of_titles_added ~ year_added.

A screenshot of a computer code AI-generated content may be incorrect.

Results:

           The regression analysis revealed a strong upward trend in the number of titles added to Netflix over time. The slope estimate was 169.05, indicating that Netflix added approximately 169 more titles each year on average. This relationship was statistically significant, with a p-value of 5.76e-05. The model also produced an R² value of 0.7531, meaning that about 75% of the variation in the number of titles added per year is explained by the year itself, demonstrating a substantial and meaningful trend in Netflix’s content growth.

A graph with a line AI-generated content may be incorrect.

Interpretation:

          There is a strong and statistically significant upward trend in the number of titles added to Netflix each year (p < 0.001). The slope estimate (169.05) indicates that, on average, Netflix added about 169 additional titles per year. The R² of 0.7531 shows that 75% of the year-to-year change in title additions is explained by the year itself, demonstrating a major expansion in Netflix’s library. 

Final Summary

            This project analyzed 8,807 Netflix titles to examine how content characteristics differ across type, rating, and time. A t-test showed a significant difference in release years between movies and TV shows, with TV shows being more recent on average (p < 0.001). An ANOVA test revealed strong differences in movie duration across rating categories (p < 0.001), indicating that content length is related to audience rating. A linear regression model demonstrated a significant upward trend in the number of titles added to Netflix each year (p < 0.001, R² = 0.75). These findings suggest that Netflix’s catalog has expanded rapidly in recent years, with notable differences in content type and rating patterns.

References

Shivam B. (n.d.). Netflix Shows Dataset. Kaggle. Retrieved from https://www.kaggle.com/datasets/shivamb/netflix-shows?resource=download

Comments

Post a Comment

Popular posts from this blog

Descriptive Statistics: Comparing Two Data Sets in R

Image
                         I worked with two small data sets, each with 7 numbers, and looked at things like mean, median, mode, range, and standard deviation. Both sets had the same spread, same range (8), IQR (3), variance (8.33), and standard deviation (2.89), which means they’re equally spread out. But their averages were really different. Set #1 had a lower mean (4) and median (3), and most of its values were on the low end, with one big outlier (10) that pushed the average up. Set #2 had a higher mean (14) and median (13), and its numbers were more balanced around the middle. This activity helped me see how two data sets can have the same variability but totally different patterns based on where the values are center
Read more

The Art of Programming Assignment

Image
 For this assignment, I created a vector in R called assignment2 using the numbers 6, 18, 14, 22, 27, 17, 22, 20, and 22. A vector in R is simply a collection of values stored together under one name, making it easy to perform calculations on the entire set. Next, I used the custom function myMean , which works by adding all the numbers in the vector with sum(assignment) and then dividing that total by the number of elements using length(assignment). Running the function gave me a result of 18.67 , which represents the average of the dataset. This shows how we can create our own function in R to calculate something as common as the mean, even though R already has a built-in mean ( ) function.
Read more

Understanding Regression Models: Predicting, Analyzing, and Interpreting Data in R

Image
 In this assignment, I explored simple and multiple linear regression using different datasets in R. Regression basically helps us understand how one variable affects another, and it can also be used to make predictions. For the first example, I looked at a dataset with x as the predictor and y as the response. Using the linear model Y = α + β X + ϵ Y = \alpha + \beta X + \epsilon Y = α + βX + ϵ , I used R’s lm() function to find the intercept and slope. The slope told me how much y changes for each one-unit change in x , while the intercept gave the value of y when x = 0 . This simple model makes it easy to see the trend in the data and predict y for any given x . Next, I worked with the faithful dataset to predict eruption durations based on the waiting time since the last eruption. Again, the regression model helped me quantify the relationship. Using lm(eruptions ~ waiting, data=faithful) and the predict() function, I estimated the discharge duration for a waiting ti...
Read more