Stata Tutorial: Correcting Autocorrelated Errors in OLS

A simple walk-through of how to use three options for dealing with auto-correlated errors in a simple OLS framework: first-difference, generalized difference, and robust standard error models are shown in Stata, and briefly explained.
Testing for Autocorrelation - Part 1:
• Stata Tutorial: Testin...
Link to "Gentle Introduction to Stata"
www.amazon.com...
Link to the excellent Introduction to Econometrics Textbook by AH Studenmund:
www.amazon.com...
Link to Jeffrey Wooldridge Introductory Econometrics Textbook:
www.amazon.com...
My Twitter is:
/ michaelrjonas
My Google Scholar Page:
scholar.google...
ResearchGate:
www.researchga...

[@humanparaquat69]

I use this channel all the time when doing my econometrics assignments at the university of amsterdam. Thanks a lot!

[@siyavashmonajem8719]

I had a project for estimating the function of money demand in Iran, because my data were time series so I had Auto - correlation problem and I wasn't able to fix it, but when I saw this video I fixed my project. I just wanted to say thank you so much Mike! keep going :)

[@venuskaraitiana]

This is literally getting me through my last minute study session. Thank you for your step-by-step easy to follow videos!

[@chetnabhagat5472]

This is extremely helpful sir, thanks a lot! The intuitive explanations that you give side by side help in making concepts clear and easily comprehensible

[@aditya-dixit]

As always amazing video, thank you!

[@mikejonaseconometrics1886]

Thank you, Aditya - Hope you are doing well!

[@patronusstag]

So so helpful, thank you so much!

[@vitaly_p]

Very helpful! Thank you!

[@MdSalaUddin-ud7yf]

Hi, I Have tried to follow you but I can not go father because, 1) reg e2hat 1.e2hat e2hat: factor variables may not contain noninteger values r(452); And in second way, 2)Durbin-Watson statistic (original) 0.585531 Durbin-Watson statistic (transformed) 2.992524 convergence not achieved r(430); Could you please explain me, why I am facing these problem. Thanks!

[@zoozolplexOne]

Cool !!!

[@ywang1801]

Hi Mike, thanks for the video! May I ask how can we determine the number of lags when using the newey-west test?

[@JayB948]

wonderful

[@mikejonaseconometrics1886]

Thanks!

[@chrisdaniels4446]

Thanks a lot. But how do you check for autocorrelation for logistic regression and how do you solve for the autocorrelation if it exists in logistic regression

[@hangwang5688]

Thanks for the video. But how to interpret the coeficients in first-difference and generalized difference? I noted that these coeficients are different from the coeficient derived from OLS and newey-west.

[@tsegayemulisa4143]

thanks! but what about longitudinal panel data to fix such problem??

[@ladydija3228]

Please how can we run the prais winsten regression for correcting autocorrelated errors and heteroscedasticity in the same time ?

[@sujaanaryal6355]

Generalized Least Squares was a bit unclear (especially with the formula), rest looks awesome as always. :))

[@frostivied9239]

Hello, thanks for the tutorial. I wonder if this is applicable to panel data?

[@qualitya407]

HI dear Jones. the question is: which one of these methods is better for all kind of autocorrelation model such ARMA and ARIMA......??? and can you proposed me a reference to study the robust st..... model?

[@FrentescuMario]

What if after applying the First Difference Correction the r squared value is way lower, to 19% from 52%? Is the model stil Valid?

[@rabihel-habta313]

Amazing! Have onenuqestion though since we're working in a time series context why you didn't test for stationarity before regressing variables?

[@saber5543]

Hi, Thanks for the video. I came up with some questions. I am working with a time series dataset, and I have nonstationary variables, all I(1). there are also heteroskedasticity and autocorrelation issues as well as non-normal distribution in the error term. I know all of these are weird, but would you give me some hint about how I can deal with these issues? (1) how can I double-check whether my data's Autocorrelation issue is solved after using newey command? (2) What is the test and correction for autocorrelation for I(1) variables? (3) what are the steps in my case? I mean, do I need to correct the nonnormality then check the stationary? Thank you very much.

[@liangchengwengchen111]

Hey, Mike. Thanks for the video! but what happens when rho is negative e.g."-0.2"? However, it helped me to fix the negative autocorrelation from 2.3 to 1.985

[@mathiasgrunenberg443]

Helllo Sir :) I wanted to ask you how I check the autocorrelation once I used the HAC standard errors... when I write the command estat dwatson right after the newey regression... its doesnt give me any number?

[@fatihaarar5770]

Good evening sir, I want to ask you, I have estimated the panel models and I did the Breusch-Pagan LM and F test I found that the PRM aggregate regression model is suitable and when I checked the autocorrelation I found that there is autocorrelation between the errors How do I correct this autocorrelation

[@mariaelisadallaverita5434]

Thank you for the videos! They are helping me writing my final master dissertation. Although, I have a question, how can I choose the optimal lags when using Newey and West? Thank you!

[@mikejonaseconometrics1886]

You will need to establish the order of the autoregressive process for the residuals, which technically requires the Box-Jenkins specification process and the autocorrelation and partial autocorrelation functions. An approximation can be achieved by running a regression of the residual on the first two lags, and if both lags are significant, it's likely an AR(2), and a second lag should be used in the newey-west calculation.

[@mariaelisadallaverita5434]

@@mikejonaseconometrics1886 Many thanks for the help!

[@Dr_Shiny]

But the R^2 for all these solutions are very low. How to justify this Sir?

[@jovial129]

15:57