The purpose of this project was to analyze Amazon review, comparing the star ratings from reviews generated by the Vine program to reviews that were not generated by the Vine program to determine if there was a significant difference in favorability.
After filtering for helpful reviews that, there were 4,726 Vine reviews that met the filter conditions and 102,218 non-vine reviews that met those same conditions.
Below is the analysis for the in_vine data set.
Below is the analysis for the not_in_vine data set.
There were 1,911 Vine reviews that were 5 stars and 46,599 non-Vine reviews that were 5 stars.
40% of the Vine reviews were five stars; 46% of the non-Vine reviews were 5 stars.
Given that the 5 star reviews generated by the Vine program are actually lower than the reviews generated outside the program, there does not appear to be a bias for favorable reviews apparent in the Vine data. An additional analysis could make use of machine learning to perform sentimental analysis to get a broader sense of how favorability is distributed across sets. We could also broaden the analysis by looking at 4 and 5 star reviews as positive, 3 stars as neutral, and 1 and 2 stars as negative to get a fuller picture of the favorability distribution.