2010LectureTiming.docx by dandanhuanghuang


									                                            2010 Lecture Timing and Content Notes

Unit 1: Overview

Lecture 1, Orientation: Timing perfect for N=24. Did not describe research interests.

Lecture 2, Sampling Distributions: Barely fit into one lecture- consider shortening slightly

        Add slide to show distribution of MAS scores (density plot) with the 39 highlight. Show the same point in the
        normal distribution

        Add slide to show the z-test example in graphic form too (highlight 18.6) in sampling distribution (normal) and
        standard normal distribution

General: Read sampling distribution chapter from FSU text vs. pdf from web. Consider replacing and making one a
supplement. Also consider these chapters from Judd

Judd, McClelland, & Ryan (2009). Simple models: Definitions of error and parameter estimates. In Data Analysis: A
Model Comparison Approach.

Judd, McClelland, & Ryan (2009). Simple models: Models of errors and sampling distributions. In Data Analysis: A Model
Comparison Approach.

Do a bit more reading on unbiased estimators. Maybe its not mean of SD = pop mean (e.g., consider positive bias in r

Bivariate Regression (NOTE: Completed 4 slides in previous lecture)

Lecture 1: slides 5 – 31

Lecture 2:

Update descriptive statistics on slide 38

Multiple Regression

Lecture 1: Slides 1-23 (no changes)

Lecture 2: Slides 24-48

        Update caseanalysis.Plot to handle bivariate scatter plots. Maybe add simple correlations. Turn recording on
        for figures

Lecture 3: Slides 49-68

Lecture 4: Slides 69 – 90
Lecture 5: Slides 91 - 112

Lecture 6: Slides 113 – 122


          Add slide with help on power. f2 from pdf

          Need to shorten this unit or break apart?. Consider removing transformations of proportions? How else can
          transforms be shortened?

          Need to add back the general info on causal models. And suppression?

          Add slide on how to extract case analysis indices or modify function to return that for identified cases

          Define pr and sr in terms of SSR, SSE, SST. Write new functions that cal each based on SS

Categorical Variables

Lecture 1: 1-26

          Consider renaming control group to something that doesn’t include control

          Consider renaming group to PatientGroup to make it more descriptive

          Need a graph of means with error bars that are SE for b0

Lecture 2: 26 – 50 (few slides not covered)

          Simplify description of coefficients for POC. Way too complicated. Consider setting them up so they can do
          homework question with non set U set V approach?

          Consider if they really need effects coding?

          Add slide to demonstrate test of orthogonal with more than 2 contrasts

          Update figures for parsing variance to be R figs

Additive Models

Lecture 1: 1-14 + handful of remaining slides from last unit

Lecture 2: slides 14 – 35 with 10 minute review of slides 1-14 b/c of long break over spring break

Lecture 3: Finished Additive models and completed 13 slides on interactive models.

Consider adding back in the slides that talk about residual approach to understanding ANCOVA (Cuts slides in folder).

Interactive models
Lecture 1: 1-13 (after having to complete 10 slides of Additive models). Worked well but consider changing the nature
of the examples (i.e. religion and gender on birth control seems weird). Same for abortion example.

Lecture 2: 14 – 30

Lecture 3: 31 – 57: The second example needs some work. Pretty redundant slides on figuring out what all coefficients
mean in the various models. Could also have better slides on defining how each effect is determined from the
interaction contrast. See non-linear slides as a model?

Lecture 4:

Non-linear models

Two lectures. (1.5 really)

Causal models

1.5 lectures

Move to Unit 4. Further emphasize the need to always interpret the b’s as controlling for other variables.

Consider redoing suppression comparing raw bs from bivariate vs. multiple regression models.

Introduce bootstrapping for indirect test for mediation


Change sticky tar to two examples – murders as outcome or dirty shoes as outcome

General to do for next time

Update formulas on slides using LaTreX or Mathplot in R

Consider really reducing the standard score model coverage in lecture. Remove info about Beta but talk about
standardizing variables as an attractive transformation for ease of interpretation

Work on standard naming conventions for all files, variables, objects, etc. Present in first class as rule to follow. Present
rules for variable naming from matlab course website?

Use “regressors” to describe terms in model. This can be consistent for all mechanical variables and also for
transformed variables.

Carefully review clarity of midterm exam questions. Be obsessive

Increase number of sample midterm questions

To top