UM's American's Changing Lives (ACL) survey series is an ongoing, national longitudinal study that covers a wide range of sociological, psychological, mental, and physical health items. I chose this dataset for its depth and breadth, because I wanted to answer a nebulous question: what factors contribute to the onset of depression? Having recently read Johann Hari's book, Lost Connections, I wanted to see if data could back up his claim that depression isn't a biochemical issue, but instead triggered by the isolation caused by social issues, specifically fractured community structure and Western society's increasing conflation of status with purpose.
After weeks of struggling with R and factor data types, I was able to produce some "descriptive statistics" charts for the midterm. Selections are reproduced below:
After the midterm, I spoke with Yoav about next steps. He introduced me to practice of "bootstrapping", telling me to:
The survey questions seemed to be coded pretty consistently, with lower factor levels corresponding to "positive" responses and increasing to "negative" responses. However, the factor levels ranged from 2 to something like 48, so Yoav taught me how to convert them all to the same scale. As for the categories, I chose the following to look at:
With all this prepped, it was time to write some code. Being a R newbie, I didn't know any fancy tricks, so my first stab looked like this:
Unsurprisingly, R didn't like this and chose to act erratically instead... so I took Kaushik up on his offer for R help. After sitting down with him for barely more than an hour, he made more progress than I did the entire semester. Guess which one of us went to art school?
Basically, what he had helped me do was to calculate the mean scores for each category for the 1994 sample, from Wave 1-3, then organized it into a tidy data frame:
So after Kaushik bailed me out of the hell I made for myself, I functionalized what he'd done so that I could apply it 100 more random samples. Being sick and tired of my own ineffectiveness with R at this point, I exported the results to csv's for visualization below: