.png)
120 – Non-Negative Matrix Factorization
Using Factor Analysis and Quantitative Content Analysis to Detect Themes in Media Texts: A Comparison of Pre- and Post-9/11 Song Lyrics
Brenda Osuna
University of Southern California
Reagan Rose
University of Southern California
Media, social networking, and communications data is of vital importance to every sector of business, industry, government, and academia. Research studying the effects of media in these industries impacts public policy and shapes decision-making in personal and business spheres. Creative application of statistical methods can potentially reduce bias and increase precision in measurement of media content. Osuna (2011) analyzed a sample of music lyrics (N=300) using Linguistic Inquiry Word Count and principle components analysis. Factor scores for each lyric were obtained on the resulting dimensions. Each lyric was measured on a continuous scale for thematic content. The validity of the model was tested using Latent Semantic Analysis. Here, the utility of this method is demonstrated through an analysis of samples of pre- and post-9/11 music lyrics to obtain factor scores. Then, using a t-test for each factor dimension, the lyrics are examined for differences. Results indicate that the method is worthy of more exploration and validity testing. We recommend further study of methodology and measurement techniques using statistical methods applied to media content.