Stylistic (documentation) changes about lost-stats.github.io

[NickCH-K]

I'll work on changing the way package names are referred to.

Thumbs-up from me on grandchild pages. I wasn't able to get it to work when I was porting everything over (thus dataset combination no longer being in a grandchild format) so I may need some guidance here (and on the Contributions page).

I'm not so sure about moving Implementations. The concern that I have is that Implementations has the potential to end up being very long if lots of languages are added, and most people probably won't scroll to the bottom for the rest. Keep in Mind and Also Consider should be pretty short, and they will include some "best-practices" information we probably want to be a little more difficult to skip.

[khwilson]

From an interface perspective for implementations, you can do a simple accordion to hide the language you don't want. In a future version, it might make sense to have a "preferred language" so that your preferred language is always open by default.

You could also do a tabbed interface.

Just spitballing :)

[grantmcdermott]

Nick, here's an example of grandchild page: https://lost-stats.github.io/Model_Estimation/Nonstandard_Errors/HC/hc_se.html

The YAML actually stays pretty much the same. Even the optional grand_parent: GRAND_PARENT TITLE field is just for full indexing on the LHS. Everything will still work without it as long as you've correctly specified a parent...

Speaking of which, you might have run into problems previously because you forgot to change the has_children field of the parent's YAML to "true"? (See here.)

[NickCH-K]

Strange, I think I was doing all of that, including has_children, but all the grandchildren pages were showing up as top-level along with Model Estimation, etc. Not sure what I was doing but it seems to be working now!

[grantmcdermott]

Okay, here's my proposed tree structure for the Model estimation section (incl. some example pages that have yet to be added):

• Linear models
  • Ordinary Least Squares (Linear Regression)
  • Fixed Effects
  • Random Effects
• Generalised models
  • Probit Model
  • Logit Model
  • McFadden's Choice Model (Alternative-Specific Conditional Logit)
  • Heckman Correction Model
• Nonstandard Errors
  • HC
  • HAC
  • Cluster-robust SEs

Thoughts?

[NickCH-K]

I'm not sure I like the "Generalized Models" title - seems vague. Perhaps Non-Linear Models (to pair with Linear Models)? Heckman would have to move. Perhaps "Multiple Equation Models" which would also end up getting IV, SEM, SUR, etc.