Generating data in Week 2, statistical frameworks in Week 4

[drphilmarshall]

@abmantz This is the branch where I'll rework our straight line example and approaches to inference chunks into new lessons for 2019 week 4. Basic plan could be:

• Introduce cepheids dataset and do example straight line Bayesian inference, first on gradient-intercept parameter grid and then by MCMC, as a demo.
• Skip "cost of gridding" exercises as we'll have done that in week 2 homework
• Maximum likelihood estimators, use to get straight line fit parameters
• Uncertainties in frequentist estimators, wording of confidence intervals
• Freq vs Bayes discussion

In this scheme, most of the new stuff needed is the MCMC demo, and thought about appropriate active learning activities. Probably the straight line demo plus exercises could be all of Week 4 Monday's session; I would not expect all of the above to take more than 3x40 min chunks, so with project speed dating as the tutorial in week 4, that'd be the whole week accounted for and no more. That would leave us with 40 mins on Week 5 Monday to fill. Thoughts?

PS. I started a lessons/README.rst file to provide links to notebooks with badges showing which are passing and which are failing (which would be generated by travis running the DESC beavis script). Useful to have such a listing, even without the CIT badges?

[abmantz]

Some immediate thoughts

1. Please make sure the cepheids introduction and especially the code is consistent with the new week 5 tutorial. It looks like I eliminated the separate code file as being unnecessary and obfuscatory (in the tutorial), although maybe we would want to hide some of the code in lecture. I'm tempted to say instead that we should just find a different example for the "frameworks" lecture to avoid repetition (and having to choose to show or hide a bunch of data munging in lecture). This isn't a place where we absolutely need to use real data, and likewise there is no longer a need to introduce the cepheid data here, since the tutorial does so.

2. I'd like to avoid redundancy between lessons/README and doc/Schedule, which is why the current lessons/README.md just links to the schedule. The badges could go in the schedule, although I personally don't see the need to use either them or travis.

3. What is the new MCMC demo you're referring to?

4. I'm not going to be concerned about whether we've allocated too much time for this until it's revamped. Everything always takes longer than we think.

[drphilmarshall]

@abmantz For reasons that made sense at the time I've included the data generation lesson in Week 2 in this PR; tutorial and HW PR is on the way. You might like to look at both these PRs together - Ill cross-reference them when the other one is up.