## My visit to the John Jost Lab at NYU--comment/question interludes plus a tiny portion of asymmetry

One of the really fun things I did recently was to give a presentation to John Jost's lab at NYU.

My talk was similar to one I've been doing recently on the respective roles of "science comprehension" and "science curiosity" as "epistemic virtues" in a democratic society (slides here).

The one innovation in the presentation was the introduction of the designated "comment/question interludes":

This is a good device, I think, where the workshop uses an "interactive" format that allows questions throughout the talk. I don't like that format much. The continuous, self-propagating flow of queries can easily wreck the organization of the speaker's remarks and block him or her from even reaching the most important material (economics workshops are infamous for this).

But sprinkled abundantly and strategically through the presentation, and announced at the start of the talk, the "interludes" kept the show going in a basically linear direction and at an appropriately brisk speed.

Or at least so it seemed!

As you likely gathered, the talk was *not* about the "asymmetry thesis."

Nevertheless, the issue came up a few times, mainly in relation to slides like these:

These pdd's were derived from logistic regression models that had left_right political orientations & (in the case of the "Motivated Numeracy") appropriate cross-product interaction terms.

I used to believe that models such as those aren't appropriate for testing the "asymmetry thesis," b/c they assume a "linear" impact for right/left political outlooks in log-transformed space.

But now I'm less sure, in part b/c of papers that convincingly contend that non-linear regression models (e.g., logistic & ordered logistic), & in particular ones that include cross-product interaction terms, support inferences about real-world non-linearities once the log-transformed values are transformed *back* into predicted probabilities. Indeed, the papers in question insist that such re-transformations, accompanied w/ appropriate graphic illustrations, are the *only* thing that support such inferences--the regression coefficients in such models are not informative on their own (Karaca‐Mandic, Norton, & Dowd, 2012; Ai & Norton, 2003; Greene, 2010; Powers, 2005; Mitchell & Chen, 2005)....

Gratifyingly, this point was not lost on the workshop audience, all of whose members, Jost included, recognized that determining which methods to use to investigate asserted "asymmetries" in politically motivated reasoning is a complicated issue.

Face-to-face can often generate progress that dueling studies elide.... And vice versa!

*References*

Ai, C. & Norton, E.C. Interaction terms in logit and probit models. Econ Lett 80, 123-129 (2003).

Greene, W. Testing hypotheses about interaction terms in nonlinear models. Econ Lett 107, 291-296 (2010).

Karaca‐Mandic, P., Norton, E.C. & Dowd, B. Interaction terms in nonlinear models. Health services research 47, 255-274 (2012).

Mitchell, M.N. & Chen, X. Visualizing main effects and interactions for binary logit models. Stata Journal 5, 64-82 (2005).

Powers, E.A. Interpreting logit regressions with interaction terms: an application to the management turnover literature. Journal of Corporate Finance 11, 504-522 (2005).

## Reader Comments (1)

link drop:

https://www.businessinsider.com/psychological-differences-between-conservatives-and-liberals-2018-2