Add pre-test post-test non equivalent group design details to the Knowledge Base

[drbenvincent]

We were previously missing this. And I was also looking at the ANCOVA page and realised it needed some work. This PR lays some groundwork for improving the ANCOVA docs/page, but doesn't touch it yet. So far, we just add info into the knowledge base docs page.

Note: This PR also updates the dependency from daft to daft-pgm. This is to keep up with a change of the package name on pypi.

Very happy to get any corrections, or feedback to improve clarity or accuracy. I think that's particularly important in this case because I made up this DAG myself - it doesn't feature in any of the (many) causal inference books I have. There's a chance a DAG has been put together in a published paper that I've not seen, so happy to hear about that if you know of such a paper?

---

📚 Documentation preview 📚: https://causalpy--517.org.readthedocs.build/en/517/

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.19%. Comparing base (09adfd7) to head (6d2e5a7).
⚠️ Report is 9 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #517   +/-   ##
=======================================
  Coverage   95.19%   95.19%           
=======================================
  Files          28       28           
  Lines        2457     2457           
=======================================
  Hits         2339     2339           
  Misses        118      118           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[drbenvincent]

Actually, is it worth mentioning more backdoor paths:

• $T \leftarrow Y_0 \leftarrow U \to Y_1$. But blocked by observing $Y_0$
• $T \leftarrow X \rightarrow Y_0 \leftarrow U \rightarrow Y_1$, but blocked by observing $Y_0$ and $X$
• $T \leftarrow Y_0 \leftarrow X \rightarrow Y_1$, but blocked by observing $Y_0$ and $X$

[drbenvincent]

Maybe I need not mention ANCOVA specifically. Just say 'linear model'.

[drbenvincent]

Just noticed... what I currently have listed as backdoor path 3 is of course easily closed by observing $Y_0$

[drbenvincent]

• DAGS -> DAGs

[NathanielF]

DAG looks good to me and Daggity:

Assume Us are latent but non-influencing

but breaks if as you say u2 can't be ignored

[NathanielF]

The ability to discount or ignore U2 is crucial.

Reply via ReviewNB

[NathanielF]

It's a case of conditional ignorability....

[NathanielF]

I wonder if it's worth pivoting out of this example and saying something now about identification in general in causal inference. This ANCOVA is a strong example... to jump off and discuss how all the strategies above are identification strategies.

Reply via ReviewNB

[NathanielF]

Looks good to me. Added a note about maybe generalising your discussion of identification - leaping off the ANCOVA example to reflect on the set of DAGs above... but just a suggestion. Not blocking for this PR...