Last year we discussed the exciting new introductory social-science statistics textbook by Elena Llaudet and Kosuke Imai.
Since then, Llaudet has created a website with tons of materials for instructors.
This is the book that I plan to teach from, next time I teach introductory statistics. As it is, I recommend it as a reference for students in more advanced classes such as Applied Regression and Causal Inference, if they want a clean refresher from first principles.
This is awesome, thank you Elena and Kosuke for making these available and Andrew for sharing! Bought the textbook a while ago, looking forward to going over these materials.
In their lecture on NHST, they have (and repeat) this: “When p-value >5%: we will fail to reject the null hypothesis and conclude the average causal effect is not statistically significant (likely to be zero at the population level)”.
oh no. The slides do look good overall though.
Another thing that bothered me in the lecture 20 slides was that the authors missed a chance to point out the Type M/S error issue, which can have really dramatic consequences with this whole $p<0.05$ ritual. Maybe they point it out later.
I would suggest not using dropbox for releasing the slides but rather a standard web page. Also, it could be nicer to have an R package with all the data in it; this could still be done. I see the advantage of showing the data by chapter as well.
I am teaching intro stats this semester and I added this book as optional reading, although if Andrew is right that nobody reads books any more then that may be pointless (at least people might read slides?).
Shravan:
I don’t like having an R package with all the data because it makes the data a bit too convenient. I prefer having the data on a convenient github page, and then students can download the data or just read them directly from the web with code such as
earnings <- read.csv("https://raw.githubusercontent.com/avehtari/ROS-Examples/master/Earnings/data/earnings.csv")Hi Russ,
Thank you so much for pointing this out! This an error in my lecture slides, not in the book, and I take full responsibility. I was trying to use clearer language for my students and, embarrassingly, I ended up implying that we accept the null when we fail to reject it. I have already fixed the slides instructors can preview on my personal website. More importantly, I will fix the raw files instructors get through the official PUP site when they teach the book in their courses.
Thank you again for taking the time and care to look closely at my slides!
Elena
In their intro lecture on observational studies, they have this: “We must identify all potential confounding variables and statistically control for them using a multiple linear regression model”.
Russ:
Thanks for pointing these out. The authors should definitely fix these errors in the next edition. It’s often a problem with introductory statistics when covering complicated topics, that the goal of making things simple and clear runs into the goal of getting it right.
This is a really great book and I’ve already used it to underpin the logic of my undergraduate course this year. Yet, I have a challenge, and I would love to hear your – and others’ – suggestions. I’m supposed to teach introductory quantitative methods as part of a larger methodology module. I have 5 weeks available in total, with 4 hours of weekly contact time (split across a lecture and IT labs). The students are UK undergraduate sociology students in their 2nd year (on a 3-year UG course), so most of them have not studied any maths since the age of 16. This is the first and only quantitative methods “course” on their degrees. What would you focus those 5 weeks (20 hours in total) on?
My interdisciplinary postgrad course is less of a headache; “Regression and Other Stories” as background reading and published research articles for partial replication exercises.
Hi Chris,
That’s a tight schedule, but one could, in principle, teach my course in five weeks, especially if you don’t have to administer an exam during that time.
My course is meant for students with no prior knowledge of statistics or coding and only minimal knowledge of math, so it skips the more advanced-level material in our book.
My course also spends a lot more time with causal inference than with survey research and predictive models, because I think it’s important for my students to walk away understanding the key elements and concepts related to causal studies and knowing how to evaluate them (even if it’s just at a basic level).
Here is the schedule I would recommend for your five weeks. It shows which of my lectures I would teach in each week. (You can find my syllabus and lecture slides here: bit.ly/teaching_dss)
WEEK #1
Lecture 2. Introduction to R and RStudio
Lecture 3. Observations and Variables
Lecture 4. Computing and Interpreting Means
WEEK #2
Lecture 5. Causal Effects and Randomized Experiments
Lecture 6. Does Social Pressure Affect Turnout?
Lecture 7. Survey Research and Exploring One Variable at a Time
WEEK #3
Lecture 9. Exploring the Relationship Between Two Variables
Lecture 10. Predicting Non-Binary Outcomes with Linear Regression
Lecture 11. Predicting Binary Outcomes with Linear Regression
WEEK #4
Lecture 13. Estimating Causal Effects with Observational Data and the Problem of Confounders
Lecture 14. Controlling for Confounders Using Multiple Linear Regression
Lecture 15. Internal and External Validity
WEEK #5
Lecture 19. Probability
Lecture 20. Hypothesis Testing with Coefficients
Plus one additional lecture from the following three:
Lecture 21. Do Small Classes Increase Probability of Graduating
Lecture 22. Do Women Promote Different Policies Than Men?
Lecture 23. Does Social Pressure Affect
I hope this is helpful!
For whatever strange reason, the Website link (for “tons of materials”) now redirects to an unrelated(?) professor’s vanity site. The link one wants is here:
https://ellaudet.github.io/dss_instructor_resources/
Link updated; thanks. I’m still puzzling over what happened with that earlier link; I’ll devote a future topic to the question.