![Doug McKee](/img/default-banner.jpg)
- 12
- 997 754
Doug McKee
Приєднався 17 січ 2015
Usually econometrics is taught math first: "Here's a nice statistical model, here are its formal properties, and here are a few applications of it." When I teach, I start with a substantive question and then ask if we can answer it with the methods already know. If not, we develop a new method that can. This approach works a lot better for people that care more about the real world than pure math.
The videos posted here don't all individually follow this approach, but they can be dropped into a class that does. Think of this is a place for regular people to learn some statistics and econometrics.
You may also be interested in my blog: teachbetter.co
The videos posted here don't all individually follow this approach, but they can be dropped into a class that does. Think of this is a place for regular people to learn some statistics and econometrics.
You may also be interested in my blog: teachbetter.co
Introduction to Randomized Experiments
This is an introduction to randomized experiments that I recorded for my probability and stats students when we had a big snow storm and classes were canceled. It draws heavily from a lecture Lanier Benkard gave when he was at Yale many years ago--Thanks Lanier!
Переглядів: 1 357
Відео
iPad Unleashed
Переглядів 2,1 тис.6 років тому
Teaching (or presenting) with an iPad Pro with George Orlov and Doug McKee. Learn more at teachbetter.co/
Combining data files in Stata
Переглядів 47 тис.7 років тому
In this short video I explain how to combine Stata data files and why you would ever want to do such a thing. Specifically, I show how to use Stata's append and merge commands.
Fisher's Exact Test
Переглядів 35 тис.7 років тому
Suppose you want to test whether multiple samples of a categorical variable come from the same distribution. The Chi-Square Test is great if your samples are large enough, but if they aren't, you might be able to use the Fisher's Exact Test. In this video I talk about where this test comes from and derive it for the simplest case: Two groups and a variable with two possible values.
Doing Chi-Square Tests and Fisher’s Exact Tests in Stata
Переглядів 13 тис.7 років тому
Chi-Square tests and Fisher's Exact tests are very powerful ways of determining whether whether a sample comes from a particular distribution or if multiple samples come from the same distribution. In this video I show you how to implement these tests in Stata.
An intuitive introduction to Propensity Score Matching
Переглядів 212 тис.9 років тому
Propensity score matching is a common technique used to estimate the effects of a treatment or program when you don't have a randomized controlled experiment. In particular, it's used when you have observational data that includes pre-program characteristics that determine whether or not each individual received the treatment. In this video, I work through a simple example of how it works and g...
An Intuitive Introduction to the Multinomial Logit
Переглядів 133 тис.9 років тому
This hour long video explains what the multinomial logit model is and why you might want to use it. I also explain how to interpret coefficients and how to estimate it in Stata. At the end, you are encouraged to use Stata to analyze some data on alligators: teachbetter.co/assets/alligators.dta Intended audience: Folks who have had some exposure to linear regression models, but want to learn mor...
An intuitive introduction to Regression Discontinuity
Переглядів 110 тис.9 років тому
When the circumstances are right, regression discontinuity can be an excellent way to extract causal estimates from observational data. In this video I give you a prototypical situation where RD is applicable and explain how it works. I also describe situations where the method fails and say a few words about fuzzy discontinuities. Intended audience: Folks who have had some exposure to linear r...
An intuitive introduction to Difference-in-Differences
Переглядів 233 тис.9 років тому
Difference-in-Differences is one of the most widely applied methods for estimating causal effects of programs when the program was not implemented as a randomized controlled trial. In this video I describe the situations where the method is applicable and give you the intuition behind it. I also explain how and why you might want to use regression to estimate diff-in-diff effects. Throughout, I...
Evaluating Social Policy in Latin America
Переглядів 8449 років тому
I gave this talk on January 23, 2015 to high school students attending the Yale Model United Nations. Latin America is a diverse and growing region that has been at the forefront of experimentation with innovative social policies. I presented five important examples: School vouchers in Colombia, supplementing children’s diets in Guatemala, health insurance in Costa Rica, paying mothers for keep...
An intuitive introduction to Instrumental Variables
Переглядів 88 тис.9 років тому
An intuitive introduction to instrumental variables and two stage least squares I teach an advanced undergraduate seminar on the economics of human capital where most of my students have taken just one econometrics or statistics course, but we have to read research articles that use more advanced methods. This video gives my students just enough about instrumental variables to read, interpret, ...
How to interpret regression tables
Переглядів 122 тис.9 років тому
This video is for students who have had some exposure to regression methods, but need a refresher on how to interpret regression tables.
Really good
Thanks a lot bro 🎉
Why is exp squared added to the equation ? Please someone to explain
In this EX, are y-scores the post-scores or the pre-post differences? I`m guessing just post scores? Thanks for clarifying!
😃 What an interesting score.
Amazing Video Prof!!
Thanks for your videos. I have a problem with merge I won't to merge 1:1 sam key variable . At all I have 7 dataset I like to merge. I get error says variable _merge already defined stata. How can we solve that problem?
# Clear all rm(list = ls()) cat("\f") library(dplyr) # Needed for %>% to work # This R program follows the example given by ua-cam.com/video/ACVyPp1Fy6Y/v-deo.html # demonstrating how to work with propensity score matching. # In the example there are 9 villages. A non governmental help organisation has built # health care centers in 4 of them where they thought it was most needed. So the treatment # "health care center" was not randomly chosen. Now we want to find the causal effect of # a new health care center on infant mortality rate. # Vectors with the data T=c(1,1,1,1,0,0,0,0,0) # Treated or not (treatment is a new health care center in the village) imrate=c(10,15,22,19,25,19,4,8,6) # infant mortality rate povrate=c(0.5, 0.6, 0.7, 0.6, 0.6, 0.5, 0.1, 0.3, 0.2) # poverty rate pcdocs=c(0.01,0.02,0.01,0.02,0.01,0.02,0.04,0.05,0.04) # pcdocs # combine the vectors into a data frame called "imdata" imdata=data.frame(T ,imrate ,povrate, pcdocs) # Do a logistic regression logistic_reg_model = glm(T ~ povrate + pcdocs , family=binomial, data=imdata) summary(logistic_reg_model) # Show the results of the logistic regression # Predict the propencity score propencity_score = predict(logistic_reg_model, type="response") imdata$PS = propencity_score # Add the propencity score to a new column in the dataframe # Create separate data frames for treated and control treated = imdata %>% filter(imdata$T==1) control = imdata %>% filter(imdata$T==0) paste("Average imrate in treatment group", mean(treated$imrate)) paste("Average imrate in control group", mean(control$imrate)) paste("Effect on imrate if treatment is compared to control", mean(treated$imrate)-mean(control$imrate)) # Last line show print 4.1, meaning that a health care center should increase morality rate by 4.1. # This is of course wrong, we are not estimating the causal effect because we are not # taken the pre-treatment characteristics into account. # Now instead create a new control group consisting of villages with as similar pre-treatment # characteristics as possible by using the method of precocity score matching. # Now do the matching in a "manual way" by looping though the treated villages and then find the # row with the closest propensity score among the not treated villages. for ( i in 1:nrow(treated) ) { min=Inf # set the minimum to positive infinity for ( j in 1:nrow(control) ) { if ( abs(treated[i, "PS"] - control[j, "PS"]) < min ) { # We have now found a row with less difference in propensity score, so # update the match to this row min=abs(treated[i, "PS"] - control[j, "PS"]) treated[i, "match"]=nrow(treated)+j } } } treated # List the data so we can see the matching # Finally calculate the average infant mortality rate among the treated # and compare it with the average infant mortality rate among the matched control villages imrate_sum_treated=0 imrate_sum_matched_control=0 for ( i in 1:nrow(treated) ) { imrate_sum_treated=imrate_sum_treated+treated[i,"imrate"] imrate_sum_matched_control=imrate_sum_matched_control + imdata[ treated[i,"match"] ,"imrate"] } avg_imrate_treated=imrate_sum_treated/nrow(treated) avg_matched_control=imrate_sum_matched_control/nrow(treated) diff_treated_vs_matched_control=avg_imrate_treated-avg_matched_control paste("avg_imrate_treated =",avg_imrate_treated) paste("avg_matched_control =",avg_matched_control) # should now display -7, meaning that the effect of building a health care center i # a reduction of 7 in infant mortality rate paste("diff_treated_vs_matched_control =",diff_treated_vs_matched_control)
Why does every tutorial I look up for this test not actually compute the answer. I don't know what "n choose m" means (or whatever you're saying). Plug in the values and show me the math!
mi problem is: how I do interpretate the new dataset generated after PSM? how do I create a table showing percentages of each categorical covariate I've chosen for matching?
How did you summarize the infant mortality rate lowering 7 deaths per 1000?was 1000 your sample population among treated and non treated infants??
This is great! Thank you so much
wish me luck boys, tomorrow final boss of my educational career, exam in econometrics. thanks a lot mr. McKee, your videos are so helpful in getting that helicopter view of these concepts. Good luck to anyone still struggeling, we're all gonna make it!
Great explanation!
Interesting (y)
Excellent video, thank you
Thanks , it is well explained
way more intuative than previously thought, well put thanks
Which program do you use to calculate this analysis? Are there some code packages, which can be used and upload data? Thanks!
8 years later and you are still saving lives. Thank you, sir.
Thank you! I would like to know, if there isn't a comparable group, like Rio, then how can one figure out the effect of this programme?
God, this is so good!
Awesome!
eh
why are you considering weights when calculating effect size. eg 0.25*() - 0.25*() - where did this 0.25 came from and why?
I'm speechless how a video of 20 min explained to me everything I was struggling with for a half of year Thank you a lot for making it so simple and clear. I would love more teachers to explain the basics like this.
Okay the second time watching this I finally understood. Thank you!
It sucks being stupid because I understood nothing
Thanks for the video! It is very clear, just a quick question: how did you compute in Stata the column "ps1"?
Its really helpful, but can you please tell how you calculated ps1? How can I do it in Stata?
At 16:11, can we also say that someone with a low ses, is 4.84 time more likely to like ice cream? Meaning if this data was regressed for *only* dummy variable ses=1, would that be the odds ratio?
decent vid
Amazing explanation of 2sls
Great video
thanks for important efforts
Excellent video, thank you very much! Can you maybe quickly explain how you calculated and displayed PS1 in Stata? I understand how to run the regression but I struggle to find the PS1 outputs per line, so I can actually match one line to another
This is excellent
amazingly explained! Thanks
26:30 "being female seems to have the odds of preferring chocolate to strawberry relative to male, but it's not significantly different from no effect at all". Are you suggesting that being female increases the odds of picking the chocolate flavour relative to being male? I think that it's the opposite here.
11:54 "that's the predicted probability...". Shouldn't it be a "odds" rather than "probability"?
great explanation
Sorry, hard to follow
where is the 0.25 in the equation coming from not ?
This is really helpful, thank you! :)
Thank you for this! I didn't quite understand the very last point, i.e. the difference between the points made for when DD is 'ok' (appropriate) and 'not ok'
Among the dozens of PSM videos, this stands out as simply the best. The central example, shown clearly with the intuitive elements highlighted, and the discussion at the end regarding what PSM does *not* do- are crucial and critical! One suggestion: insert a slide showing the logit regression model to really highlight where the probabilities are coming from.
Thank you!
Loads of thanks for such simple explanation
Dude. This saved me thanks :)
I found this video very helpful.