Project w/ Laura E Shummon Maass

• Complete testing and workflow (Jupyter Notebook)
• Custom functions created (Python)
• Scraping Process (Jupyter Notebook)
• Presentation (Google Slides)

• Using a dataset of reviews from Pitchfork, our goal was to answer 4 hypothesis statements about the data to provide deeper analysis of how a hypothetical artist could approach how the particular categorization and reviewer may effect their score.

• Null: There is no statistically significant difference between the reviews for Pop/R&B albums and Electronic music albums.

• Alternative: The reviews for Pop/R&B albums and Electronic albums have a statistically significant difference between how they are reviewed, either positively or negatively.

• Findings:

• The distribution of ratings for both Pop & Electronic lined up very closely:

• The sample means were very close (less than a 0.16 difference in avg rating)

• And the t-stat was below 0.37 (-0.3642)

• The p-value came to 0.7157 which far exceeded our threshold of 0.05

• We also did a bootstrap test and found:

  • Almost identical standard deviations for both samples
  • Almost identical 2.5 and 97.5 mean percentiles for both samples

• With all of the evidence above we accepted our null hypothesis that there is no statistically significant difference between the rating of Pop/R&B and Electronic albums.

• Null: In sampling the 500 worst and 500 best reviewed albums, there is no statistically significant difference in the genres each represents with regards to score.

• Alternative: The 500 worst reviewed and 500 best reviewed albums have a statistically significant difference in the genres they have.

• Findings:

• Conducted an ANOVA test on the 500 best & worst reviewed albums.

• We found trends amoung genres in the 500 best reviewed albums, but no trends emerged in the worst reviewed.

• In the 500 best reviewed albums, pop/r&b and rap had p-values of 0.0013 and 0.026 respectively. Rock had a p-value of 0.0546

• Based on the 3 lowest p-value genres, we reject the null for pop & rap, but not for rock (since it doesn't meet our 0.05 cutoff)

  • Pop & Rap genres in the top 500 albums have a statistically significant difference in review score compared to other genres.

• Null Hypothesis: There is no statistical difference between the top album reviewers and all others.

• Alternative: Top album reviewers have a statistically significant different in their review scores.

• Findings:

• Viewed distributions of the top 5 reviewers (in terms of volume) for pop and electronic:

• 1 stood out for both genres: Ian

• Mark stood out for Rock, but only had a slight differnce for Pop and Electronic.

• Compared Ian to Joe AND compared Mark to Joe:

  • Ian had a statistically significant difference from Joe
  • Mark had a slight difference, but it was not statistically significant

• Null: There is no statistically significant difference between self-released album ratings and all others.

• Alternative: Self-released albums have a statistically significant decrease in ratings compared to other labels.

• Findings:

• The distribution of ratings between self-released albums and albums with labels is not quite as closely lined up as pop and electro ratings, however they still overlap quite well:

• The sample means had a difference in rating of 0.09

• However, the t-stat was -1.44

• Our p-value was also much lower at 0.1499. This still exceeds our threshold of 0.05

• With the evidence above, we accepted our null hypothesis that there is no statistically significant difference between self-released album ratings and albums with labels.