Soc 381 daily lecture menu

* Will be revised as we go!

* Wednesday, October 6

·        Review of HW1 answers and some issues that came up in HW1

·        Discussion of the “on linear regression” tab of the class Excel file: the slope, the intercept, the R-squared, the units, and the summary statistics.

·        Section 13, on Ordinary Least Squares regression from the “mean, variance, and simple statistics” PDF file.

·        Perhaps also section 13.1 of the PDF file, on correlation, R, and r-squared (we got partly through this and will come back to this on Monday).

* Monday, October 11

• Section 13.1 of the PDF on “means, variance and simple statistics” on adjusted R-square, F-tests for comparisons of R-squares in nested models. I will probably pull up the F table from the PDF on statistical distribution tables. [did all this]
• We will move on to the “Understanding dummy vars” sheet of the class Excel file, and then on to the “regression table” sheet. The “regression table” sheet is a big one that will take us several classes to fully digest. [we covered the “understanding dummy vars” sheet and then we skipped over the “regression table” sheet and discussed the “regression graphs and fits” sheet, showing how the linear age does not predict income, but age and age-squared together do]
• We may also pull up the research terminology web link from my web page and have a discussion about bias, error, and what kinds of bias may impact our social science analyses. [we discussed this briefly and may return to it at a later time]

* Wednesday, Oct 13

     * We will spend most of class on the “regression table” sheet in the class Excel file, talking about regressions nested within each other, about F tests comparing the R-squares of nested models, and about how the coefficients in multivariable regression relate to each other. We are also going to talk about interactions among the coefficients. [We talked about how HW3. Q1-3 of HW3 should be feasible to start on based on what we have learned so far. Q4 of HW3 requires some additional information that we will talk about next week. We covered a lot of the “regression table” sheet from the class Excel file. We did not finish our discussion of interaction terms and we did not quite get through the explanation of how we calculate the predicted age peak for earnings based on the two coefficients. We will start next Monday’s class there. We agreed that the HW3 question about propensity scores, Q5, has been dropped. We will not be talking about propensity scores in Soc 381.]

* Monday, Oct 18

              * We will pick up with interactions and squared terms (looking at the Excel file, sheet “regression table.”) Then we will talk about the “alternatives to OLS SE” page, looking at the jackknife and the bootstrap. I also expect to provide some summary discussion of HW2, after those have been graded. I also used the “regressions graphs and fits” sheet of the excel file to talk about generating fitted values and graphing them (after tabling them to get an average of fitted values for every age, as distinct from graphing all fitted values). We will also talk, briefly, about the sections “the Normal and T distributions” and “OLS as an example of Maximum Likelihood”, p. 17-18 of the PDF file “notes on the mean, the variance, and simple statistics” [we did not get to this last bit]

* Wednesday, Oct 20

              * I will start with the sections “the Normal and T distributions” and “OLS as an example of Maximum Likelihood”, p. 17-18 of the PDF file “notes on the mean, the variance, and simple statistics.” I will pick up anything left over from Monday, Oct 18, and I will introduce a brief discussion of the matrix algebra of OLS regression, see the documents “NYU” on matrix algebra [now with a few corrections in it], and see also the documents linked from my webpage right after the NYU link. There is a Youtube example of inverting a matrix by Gaussian elimination.

* Monday, October 25

              * We will start with the logarithm quiz (no names, no grades). We will pull up the “logits” sheet of the class Excel file to talk about the exponential function, the ln(x) function, the odds, the log odds, and the logit transformations. We will talk at length about the “very brief introduction to logistic regression” page 11-14 of the PDF notes on mean and variance. We looked at the “logistic regression” sheet of the class Excel sheet and talk about the different ways to present results from logistic regressions (either coefficients or odds ratios), and how the two versions relate to each other

* Wednesday, October 27

              * We will follow up on a comment from last class about plotting predicted values using marginsplot. We will talk about “some comments on the Chisquare distribution,” pages 15-16 of the PDF notes on mean and variance, and about the chisquare distribution itself. Then if there is time we will visit the “LRT test” page of the Excel file, and the “std Normal and chisquare(1)” page and the “chisquare distribution” page of the Excel file. At the end of class each student will give a brief update about how their research project is coming along.

* Monday, November 1

              * We will start with the “LRT test” sheet of the class Excel file, specifically to look at the BIC and AIC tests that are adjustments to the LRT (Likelihood Ratio Test) that penalize models for having too many terms. Then we will progress to mediation tests, which draws on a new document “mediation tests” linked on my website.

* Wednesday, November 3

              * For Wednesday there is a reading assignment: the debate between Jasso and Kahn and Udry about outliers and their influence. We are going to try to finish the discussion of mediation tests, and then spend the remainder of class working on outliers and a statistic that measures the influence of each point on the slope of the line, the DFBetas.

* Monday, November 8

              * For Monday, November 8 we are going to start by reviewing Tufte’s book the Visual Display of Quantitative Information, which you should review before class and if possible bring with you to class. We will finish up whatever loose ends there are with regard to mediation and outliers. We will also be going over some HW3 answers.

* Wednesday, November 10

              * In addition to picking up any loose ends and questions related to HW4, we will discuss the Lisse et al 2003 on Recoxefib, and the Rosenfeld 2014 JMF paper on couple longevity. We will discuss the Excel file worksheets “polling, n, prediction” and “a quick note on power.”

* Monday, November 15 (week 9)

              * We are going to talk about two papers: Winship and Radbill’s paper on weights in regression, and Cohen’s paper on Bayesian analysis. Winship, Christopher, and Larry Radbill. 1994. "Sampling Weights and Regression Analysis." Sociological Methods and Research 23:230-257 and Cohen, Jacob. 1994. "The Earth Is Round (p<.05)." American Psychologist 49 (12):997-1003. We will look at the relevant sheets in the Excel file: “A note on (Bayes) Hyp testing” and “Weights and Regression.” We are also going to spend some time reviewing each student’s progress on their project.

* Wednesday, November 17

              * Read Nate Silver’s book, The Signal and the Noise. To put Silver’s work in context, I will talk about the very last section of the Means and Variance notes, the section “On Independence” and I will discuss the “Chi2 Indep tests” sheet of the class Excel file. I will also review Silver’s versus Sam Wang’s predictions of the 2016 election. Be prepared to talk more about your project. Review of HW 4 answers. A few comments about what will be on the final exam.

* Monday, November 22-24

              * No class, Thanksgiving break.

* Monday, November 29 (week 10)

              * Student presentations (on a schedule to be assigned by Professor Rosenfeld)

* Wednesday, December 1

              * Student presentations and final exam review.

* Tuesday, December 7

              * 8:30A-11:30A, final exam, in our regular classroom.