Write HMM popularity blog post

[AlexAndorra]

The wonderful symbiosis of hidden Markov models and survey data 🤩

[AlexAndorra]

This is now ready for review @rlouf 🥳 Let me know if you have any questions

[rlouf]

It's great! I see you kept some of my jokes :sweat_smile: A couple of remarks:

• Not sure if showing p_disapprove on the first figure is really informative since we do not use this information. It's also redundant with No answer. Since you mention briefly No answer I would only keep this one.
• "For Chirac's term, it seems like the difference stems from the polling method: face-to-face approval rates seem to be much lower" -> see figure below.
• Typo: "undertand the code."
• This mistake comes from me: in the model, \mu is the latent popularity, p is the popularity we can estimate from this poll.
• Name of functions and distributions should be like \operatorname{Binomial}, text like pollsters in \text{}
• "much more variation in the data that a Binomial distribution expects" -> "much more variation in the data than a Binomial distribution can account for"
• "(yeah, I'm weird sometimes)" --> TRUTH
• The ZeroSum also makes a lot of sense for house effects, maybe add a sentence about that?

[AlexAndorra]

Thanks @rlouf, just pushed the changes!