I really like this post of Nate Silver’s. Ideal-point models and other fancy statistical techniques are fine, but I’m a big fan of using the simple, directly-interpretable summary when it makes the point.
I really like this post of Nate Silver’s. Ideal-point models and other fancy statistical techniques are fine, but I’m a big fan of using the simple, directly-interpretable summary when it makes the point.
Other things which are also very powerful, but underrated and underused, for really understanding the situation well and not being mislead, are histograms and percentiles.
There's a very strong tendency to want to look at one or a few point statistics to describe a situation, like mean and variance, and they can only tell you so much, and are often misleading, but the histogram and percentiles can give such a wealth of information and insight, and make it so much harder to be mislead by statistics.
Of course, often you'll want advanced models and techniques on top, especially with many variables, but I have found histograms and percentiles (and confidence intervals) to be very underutilized in favor of neater, simpler, point measures.