Papers of the day   All papers

On the Nuisance of Control Variables in Regression Analysis

Comments

Beyers Louw: I am excited to share my first working paper! @PHuenermund and I address why the interpretation of control variables in regression analysis should be avoided. “On the Nuisance of Control variables in Regression Analysis” http://arxiv.org/abs/2005.10314 #econometrics #causalinference /1 https://t.co/6eaPAiIXAO

16 replies, 574 likes


Jason Kerwin: This is the correct take but I’m not sure I’ve seen it made in an actual paper before. A *lot* of papers I referee and grade spend time talking about the coefficients on control variables. Don’t do this.

3 replies, 152 likes


Paul Hünermund: Because of the success of the blog post, we decided to make a citeable research note out of it: "On the Nuisance of Control Variables in Regression Analysis" (now available on: https://arxiv.org/abs/2005.10314). Please use it and save yourself a paragraph or two in your next paper! ☺️ 1/2

5 replies, 128 likes


Sebastian E. Wenz: So sad that such papers are still needed; all the more important that they get written and published.

0 replies, 28 likes


Niels Holtrop: @rlmcelreath @TaylorMcLinden @PHuenermund @Beyers_Louw have a paper on this too, aimed at an economics audience: https://twitter.com/Beyers_Louw/status/1263863628538753025?s=20

1 replies, 22 likes


Yongnam Kim: Don’t report & interpret the coefficients of control variables. Love the authors’ last sentence. “This approach will [...] save on valuable journal space”

0 replies, 21 likes


Maria Koumenta: The paper and the discussion here are worth a read! #EconTwitter #AcademicTwitter

1 replies, 15 likes


Alex Keil: The economists are catching up to us, #epitwitter! First DAGs, and now the Table 2 fallacy. All in a matter of months. Time to dust off those GRE math skills again.

0 replies, 10 likes


Alex Turner: Excellent paper 👇 Be cautious in interpreting the effects of control variables because: (1) they are likely to be subject to omitted variable bias; and (2) other control variables (and your explanatory variables of interest) are likely colliders/bad controls.

0 replies, 10 likes


Michael DeCrescenzo: This stuff really affects my stance on "accuracy/interpretability trade-off" for machine learning in academic social science. For confounders in a causal setting, regression's "interpretability" advantage is overstated. The algebra is easy to interpret, but not the substance

1 replies, 6 likes


Noah Haber: "On the Nuisance of Control Variables in Regression Analysis" Short, sweet, to the point, an excellent title, and some fun in the appendix as well. Nice work y'all, looking forward to digging in and discussing further.

1 replies, 6 likes


David Evans: “On the Nuisance of Control variables in Regression Analysis”

0 replies, 4 likes


Jodi Beggs: I looked at the title and was like hey my coworkers wrote a paper =P

0 replies, 2 likes


Katharina Hölzle: Excellent paper with some important implications against the inflation of using control variables. @CIM_Journal @JournalPIM. Thanks @PHuenermund @beyers_louw

0 replies, 1 likes


Itamar Caspi: Brilliant. I am amazed that this paper waited until 2020 to be written.

0 replies, 1 likes


Content

Found on May 22 2020 at https://arxiv.org/pdf/2005.10314.pdf

PDF content of a computer science paper: On the Nuisance of Control Variables in Regression Analysis