Nathan Favero

@nathanfavero.com

566 Followers 710 Following

soft scientist (american university), church attendee, monoglot, datum scientist, chess grandnonmaster, etc. stata, r, javascript, quickbasic. #lifelonglearner nathanfavero.com

Brittleness: A new concept for statistical modeling
Factors like small samples, multicollinearity, and outliers may not themselves break our models, but they create conditions where a break can be dramatic
When learning statistical tools for empirical analysis, many of us were...
nathanfavero.substack.com

Reflecting on Null Results
What I have been learning and doing with null results over the past year

Substack tile: 3 rules for research presentations - how to avoid some common mistakes in academic presentations

Screenshot from article: Most notably, we find that the prioritization of equity exhibits the greatest variation across respondents from different backgrounds and social identities, suggesting that the importance of this value, more than others, remains contested and context-dependent. Similarly, representation—while generally held as a high priority by the public—is considered especially important by liberal-leaning individuals. Thus, when it comes to symbolic representation (see Keiser et al., 2022), pursuit of a representative bureaucracy may enjoy particularly widespread support among liberal constituencies. These examples suggest that public administrators should acknowledge citizens’ heterogeneous

Screenshot of Table 1 from article linked to later in thread. Summary stats (means). Equity 2.982. Efficiency 2.964. Effectiveness 3.199. Representation 3.136. N = 814; Response options: Not Important at all (0), Of little importance (1), Of average importance (2), Very Important (3), Absolutely Essential (4).

Session info: Saturday, April 5. 3:20pm. Sandburg 3: 7th floor of Palmer

Excerpt of NBC article: Public servants across federal agencies now find themselves caught between President Donald Trump's executive orders and federal judges who have blocked them — at least temporarily — until the courts can more fully determine whether the president has exceeded his legal authority. They face firing by the Trump administration if they defy him and potential legal sanctions if they don't.

These officials are at the ground level of a burgeoning constitutional crisis in which Trump is claiming expansive…

Screenshot from article: Mr. Musk’s allies now aim to inject artificial intelligence tools into government systems, using them to assess contracts and recommend cuts. On Monday, Thomas Shedd, a former Tesla engineer who has been tapped to lead a technology team at G.S.A., told some staff members that he hoped to put all federal contracts into a centralized system so they could be analyzed by artificial intelligence, three people familiar with the meeting said.

AI overview after the search “what grade would a 4th grader get”. The overview says a 4th grader would get a “grade 4”

Screenshot of AI assistant responding to search “how much colder than 27 before…” The assistant says if the temperature is 27 degrees, it needs to be five degrees colder before it freezes. It also shows the subtraction: 32 - 27 =5.

Screenshot of AI assistant responding to search “how much colder than 27 before…” The assistant says if the temperature is 27 degrees, it needs to be five degrees colder before it freezes. It also shows the subtraction: 32 - 27 =5.

Screenshot of article text, which reads as follows: “Right now when we have the Department of Education, we have an entire agency where the leadership is thinking 100 percent about education, because that’s their whole mandate,” he said. “Whereas if they slice and dice this up and it goes to different agencies … now education is just going to be a very small part of the portfolio that that organization does, and I don’t think their leadership is going to have the same attentiveness to steer new initiatives.”

My coauthor Weijie wang presenting our conclusion slide

screenshot showing pdf files for slides for 10 lectures, with a preview of Lecture 1: Graphing 1 variable. folder also shows sub-folders for data files and for Powerpoint conversions.

showing files for a class data collection project (codebook, survey results template, homework instructions, policy memo assignment description, survey text, survey-based lab). a preview of the policy memo assignment description is also shown, indicating project should use data from survey that the class jointly collected. three drafts are due, and the memo should be structured with 3 section: (1) identifying a research question, (2) description of data & measure, and (3) analysis.

screenshot of code from Stata lab showing how to use if statements and recode variables

Lab 1: basic graphing & labeling (data: college binge drinking)
Lab 2: basic data processing, descriptive stats, variable transformations, and more graphing (data: Texas school districts)
Lab 3a: bivariate data description, regression; assumes Lab 2 was completed (data: Texas school districts)
Lab 3b: basic regression examples, closely tracking Ch. 3 example (data: university GPAs)
Lab 3c: an alternative/extension to 3a, but less annotation so you may want to add your own; could also replace Lab 6 (data: U.S. states)
Lab 4: bivariate data description, confidence intervals (data: COVID-19 survey)
Lab 5: data processing; doesn't relate to Ch. 5, so could easily be another week (data: NYC schools)
Lab 6: simulations; additional simulations available with the Lecture 6 code/slides (https://github.com/favero-nate/minus-the-math/tree/main/lecture_slides)
Lab 8: comparing means through t-tests, ANOVA, & regression (data: U.S. states)
Lab 9: chi-squared

Table of contents:

Preface
1 Graphical Tools for Describing Variables
2 Statistics for Describing One Variable at a Time
3 Tools for Describing the Relationship Between Two Quantitative Variables
4 Estimation
5 Probability Distributions
6 Sampling Distributions
7 Hypothesis Testing
8 Comparing Means (How a Qualitative Variable Relates to a Quantitative One)
9 Comparing Groups (How Two Qualitative Variables Relate to One Another)
10 Causality and Experimental Designs
11 Models and Uncertainty
12 Regression with Qualitative Independent Variables
13 Regression with Qualitative Dependent Variables

1. It is a true introduction, not assuming any prior training in statistics.

2. I try to minimize use of math (beyond the very elementary) in initial explanations, instead focusing on conceptual description and interpretation.

3. I introduce regression very early on (Ch. 3) so that students (especially PhD students) can quickly get started on their term papers and better understand any quantitative articles they’re reading. The treatment of regression is further built out in the final chapters (11-13). Regression is, after all, the workhorse of applied statistics for the social sciences.

4. I skip a traditional treatment of probability theory because I don’t find traditional treatments to be very useful for students interested in applied statistics. Instead, I’ve written a brief chapter (11) on the logic and practice of building models that account for uncertainty.

5. There is a bit of Stata code in the final chapter, but otherwise all examples are presented apart from any software

Oil painting of a forest of trees changing colors in the fall