Statistical Analysis of Twitch Streaming Data

This technical report presents a detailed statistical analysis of Twitch streaming data, focusing on various regression models to understand viewer engagement and follower growth. The report begins with a note on the absence of inference due to the non-random sampling of data from the top 900 Twitch users. It outlines summary statistics for key metrics such as average viewers, stream hours, and follower counts, emphasizing the right-skewed distributions of these variables. The analysis includes simple linear regressions to predict average viewers based on mean weekly stream hours, revealing a moderate R-squared value of 0.56. Additionally, the report discusses the correlation between streaming time and viewer metrics, highlighting unexpected relationships, such as a negative correlation between stream time and average viewers. The findings suggest that merely increasing streaming hours does not necessarily lead to higher viewer counts or follower gains, challenging common assumptions in the streaming community. The report concludes with visualizations and further statistical details in the appendix.