divide gov. followers by state pop.?

[soodoku]

[mkearney]

I'm thinking I should break these up even more. I could generate a plot for each with the README being one big index. Or maybe, I could do by State, Trump/POTUS, and then one for the rest of the cabinet?

The problem now is I'm having to log the counts just to have them reasonably fit on the same graph...and that makes it even harder to see variations in follower counts. This might be fixed over time (the sample is obvi too small now), but it def seems like I'm jamming too much on too few plots.

[soodoku]

Hey,

I thought about this a bit more.

There are a few ways around this:

1. You would center by the n_followers at t_1
2. Rather than log_follower count, you could do sqrt(followers) as that is a gentler parametric transform

Two more things:

1. I would label y-axis with actual counts and more scale points to convey that it is a transformed scale.
2. If plotting raw numbers, I think it is still a useful idea to divide by state_pop. for gov. so that we aren't just tracking state population.
3. The thickness of each line can be reduced and the standard format of growth curves with labeled end points can be created.

[soodoku]

There are multiple things we could learn from this data. In Congress, from what I recall, ideologically extreme leg. have more followers than more moderate members. I think it is true for R and Ds. (One of my students did that 3--4 years ago for a DS class I taught.) So adding in CF-Scores or DW-Nominate may be useful.

The over-time growth curves can be clustered. Do growth patterns look similar for R and D? etc. etc.

[soodoku]

The first graph should just be about the skew in follower counts.

A dot plot with: y-axis = pols x-axis = n_followers