If you want to predict things that can't be precisely measured (such as social anxiety, social connectedness, and self esteem) I recommend using your numerical screen time data to perform principal component analysis.
In short, principal component analysis "uses an orthogonal transformation to convert a set of observations of possibly correlated variables into a set of values of linearly uncorrelated variables called principal components."
The principal components CAN be interpreted as measurements of something you don't have a measurement for, in this case self esteem, etc... This is a pretty big stretch so if you are doing this in a rigorous academic environment then you are going to need something more to back up your claims.
Read paragraph 2.5 in this paper to see what I mean: http://www.math.montana.edu/jobo/writing/documents/barbour.pdf