Causal Analysis in Theory and Practice

January 27, 2015

Winter Greeting from the UCLA Causality Blog

Filed under: Announcement,Causal Effect — eb @ 7:34 am

This Winter greeting from UCLA Causality blog contains:
A. News items concerning causality research,
B. New postings, new problems and new solutions.

A. News items concerning causality research
A1. Reminder: The 2015 ASA “Causality in Statistics Education Award” has an early submission deadline — February 15, 2015. For details of purpose and selection criteria, see here.

A2. Vol. 3 Issue 1 of the Journal of Causal Inference (JCI) is about to appear in March 2015. The Table of content will be posted on our blog. For previous issues see here. As always, submissions are welcome on all aspects of causal analysis, especially those deemed heretical.

A3. How others view “statistical control”.
Last month, columninst Ezra Klein wrote a post about the use and abuse of statistical “controls” (or “adjustments”), especially in studies concerning racial or gender discrimination (link). His bottom line:”sometimes, you can control for too much. Sometines you end up controlling for the thing you’re trying to measure.”

Matthew Martin, writing in here echos Klein’s concern and adds to it two other flaws of improper control: confounding and selection bias. His bottom line: “be suspicious whenever a paper says “controlling for ____”. There is a good chance you can’t actually control for that.”

I am posting these two articles to stimulate discussion on whether we have done enough to educate the general public, as well as the scientific community on what modern causal analysis has to say about “statistical control”.

B. New postings, new problems and new solutions.

B1. Flowers of the First Law of Causal Inference

Our discussion with Guido Imbens on why some economists avoid graphs at all cost (link) has moved on to another question: “Why some economists refuse to benefit from the First Law” (link). I am convinced that this refusal reflects resistance to accept the fact that structural equations constitute the scientific basis for potential outcomes; it goes contrary to conventional teachings in some circles.

But resistance aside, the past two postings lay before readers two miracles of the first law, which I labeled “Flowers”. The first tells us how counterfactuals can be seen in the causal graph (link), and the second clarifies questions concerned with conditioning on post-treatment variables. (link).

B2. Causality in Logical Setting

In the past 15 years, most causality research at UCLA has focused on causal reasoning in statistical setting, attempting to infer causal parameters from statistical data. It was refreshing for me to receive a new paper from Bochman and Lifschitz on “Causality in a Logical Setting” (link). The paper reminded me of a whole body of work that has been going on in the logic-based community, where the task is to communicate causal knowledge and reason with it commonsensibly, from beliefs to interventions to counterfactuals Worth our undivided attention.

B3. At the request of many, I am posting a copy of the Epilogue of Causality (2000, 2009) which, so far was available only as a public lecture ( I am amazed to realize that there are very few things I would change in this text today, almost 20 years after the lecture was written (1996). Still, if you spot a gap, or a need for additional stories, quotes, anecdotes, ideas or personalities, please share.

B4. Dont miss our previous postings on this blog and, of course, our steady outflow of new results, here.

Some are really neat!

January 22, 2015

Flowers of the First Law of Causal Inference (2)

Flower 2 — Conditioning on post-treatment variables

In this 2nd flower of the First Law, I share with readers interesting relationships among various ways of extracting information from post-treatment variables. These relationships came up in conversations with readers, students and curious colleagues, so I will present them in a question-answers format.

Rule 2 of do-calculus does not distinguish post-treatment from pre-treatment variables. Thus, regardless of the nature of Z, it permits us to replace P (y|do(x), z) with P (y|x, z) whenever Z separates X from Y in a mutilated graph GX (i.e., the causal graph, from which arrows emanating from X are removed). How can this rule be correct, when we know that one should be careful about conditioning on a post treatment variables Z?

Example 1 Consider the simple causal chain X → Y → Z. We know that if we condition on Z (as in case control studies) selected units cease to be representative of the population, and we cannot identify the causal effect of X on Y even when X is randomized. Applying Rule-2 however we get P (y|do(x), z) = P (y|x, z). (Since X and Y are separated in the mutilated graph X Y → Z). This tells us that the causal effect of X on Y IS identifiable conditioned on Z. Something must be wrong here.

To read more, click here.

Powered by WordPress