I am absolutely amazed at how someone can be so articulate in their explanation of this topic. You have helped me run my entire analysis for my MSc dissertation. THANK YOU! GOD BLESS YOU!
Thank you very much for the video! I would like to ask should we use the created variables (e.g., variables generated with first differences) after getting rid of unit roots in our main regressions? That part is confusing me.
I cannot duplicate the result in this video. When downloading freduse SP500, the number of observations is different. So as the result is different and fuller SP500, the command of dfuller, trend regress does not give the significant result. Please check again. Thank you Mike!
Hi, Thanks for the video. I run Breusch-Pagan with chi2(1) = 19.55 Prob > chi2 = 0.0000 and a White chi2(14) = 322.87 Prob > chi2 = 0.0000 Should I reject the null hyp? So does my modell show heteroskedasticity?
It is actual very interesting presentation , may I ask you one question ! if yes How many lag difference is appropriate to under take unit root test ,given the optimal lag length is 2 .
Thanks for the video, good introduction! How do you decide on how many days of lag you test to see whether value changes? You only test 1 and 2, is this enough compared to the 2500 observations in the data set? I've read about some "rule of thumb" somewhere else...
Thanks for watching. To generate a second difference of a variable X, you can use "gen d2x=D2.X". Higher order differences can be created the same way, and the difference operator can be used from with the regress command.
I have a question please help me ; I have a export data but ı reach the trend stationary process, so can I use this data for VAR analysis? how can I transform the trend stationary process to sationary process
thanks for the video. Just a question, if we get that we reject but we have included a trend in the test, then we should detrend the variable in order to be stationary no?. Also if we include the drift, what should we do?, thank you in advance
Hi, your video is wonderful! I just have a quick question here. When i type in the command dfuller *variable*, drift lags(1) regress, the absolute value of the test statistic is greater the the absolute value of the 5% critical value. Does this mean I can say that my time series is stationary?
@@mikejonaseconometrics1886 So when we get the first difference of a nonstationary series we generally obtain a stationary one. Now we should use the stationary series which is I(1) and forget about the original nonstationary series. Is this the case?
Hi Mike, This was really helpful...Could you do one for panel unit root testing---or if you have already, direct me to it? I'm not including a lagged DV in my panel analysis, so I'm a bit confused as to whether running a unit root test on my panel is necessary. I'm also really confused about how I'm supposed to select the lag structure. I'm using a dicky-fuller test which won't let me use the AIC or BIC protocol, so I have to select a number of lags---but I don't have a predetermined number of lags that I think should be included.
Thanks for making this video. I probably shouldn't be, but I'm always a little befuddled by the difference between drift and trend, and when to include them in the model... Am I right in thinking that (assuming a positive trend in our series) both manifest themselves in an upward slope? Given the way that time enters into the difference equation at 4mins44 surely this would lead to some sort of exponential growth in the SP500 series, and I would have assumed that this would manifest itself in the time plot of the differenced SP500 as an upward trend. Your comments would be greatly appreciated!
I am absolutely amazed at how someone can be so articulate in their explanation of this topic. You have helped me run my entire analysis for my MSc dissertation. THANK YOU! GOD BLESS YOU!
For more context, I am a petroleum engineer with no economic background. Thank you!!
thank you very much! you just helped me with my finals (and i would probably use this for my thesis as well). subscribed!
Wonderful session through this tutorial. Thank you,
Helpful video! Thank you
Good content! Subbed. Could I suggest a higher resolution on the videos? Thank you for making these videos available to us.
Excellent video!
Super helpful, thank you!
Dear Mike, thanks for teaching me in particular. Could you kindly help me how I can use covariate augmented
Dicker-Fuller in Stata. Thanks.
Thanks for the nice tutorial. Please can you help me with how to do the ADF test with trend and intercept?
What if it is stationary at lower lag and not stationary at higher lag
Great video, what happens when the variable you're trying to test on is a first-difference?
Thank you sir, do you have any tutorial of conducting Phillip perron test in stata?
Not yet, will put it on the list - thank you!
Hello sir, could you please explain how to interpret the phillips perron test with Stata
Sir,would you tell me the stata command for granger causality tests in ARDL
?
Thank you very much for the video! I would like to ask should we use the created variables (e.g., variables generated with first differences) after getting rid of unit roots in our main regressions? That part is confusing me.
Yes, that is correct. Applying OLS estimation to non-stationary variables can result in "spurious" results. Glad the video was helpful!
Emeğinize sağlık hocam, teşekkürler! 👍
I cannot duplicate the result in this video. When downloading freduse SP500, the number of observations is different. So as the result is different and fuller SP500, the command of dfuller, trend regress does not give the significant result. Please check again. Thank you Mike!
Can you do unit root test for panel data?
Actually i get worried about in which situation i should select noconstant, with trend and with drift term during adf test
Hi, Thanks for the video. I run Breusch-Pagan with
chi2(1) = 19.55
Prob > chi2 = 0.0000
and a White
chi2(14) = 322.87
Prob > chi2 = 0.0000
Should I reject the null hyp? So does my modell show heteroskedasticity?
How to do this in case of a Probit model? The command "predict resid, residuals" gives an error in case of a probit model. Thanks!
Hi thank you for the great video! is there a command to export the unit-roots tests onto word or excel?
It is actual very interesting presentation , may I ask you one question ! if yes How many lag difference is appropriate to under take unit root test ,given the optimal lag length is 2 .
Hi Mike, Thank you again for the video!! I would love to watch a video based on FGLS. Do you have in mind making one anytime soon?
Hi Mike, how can I go about Unit root tests with covariates and bootstrap CADF test?
Thanks for the video, good introduction!
How do you decide on how many days of lag you test to see whether value changes? You only test 1 and 2, is this enough compared to the 2500 observations in the data set? I've read about some "rule of thumb" somewhere else...
Thank you so much
You’re very welcome!
Thank you so much.
how to do the second difference if its not yet stationary after the first ? what is the command?
Thanks for watching. To generate a second difference of a variable X, you can use "gen d2x=D2.X". Higher order differences can be created the same way, and the difference operator can be used from with the regress command.
I have a question please help me ; I have a export data but ı reach the trend stationary process, so can I use this data for VAR analysis? how can I transform the trend stationary process to sationary process
If the random errors are not white noise what could i do then?
YOU ARE AMAZING SIR! love your videos, they are so so helpful. I am going to use GARCH model so its fine for my data to be non-stationary right?
thanks for the video. Just a question, if we get that we reject but we have included a trend in the test, then we should detrend the variable in order to be stationary no?. Also if we include the drift, what should we do?, thank you in advance
Hi Mike, thanks for the video!!! I've a request, how do you test for structural breaks in time serie data using Bai and perron test in stata?
Hi Karake, did you manage to learn how to do that in Stata? I have same the question too :)
Hi, your video is wonderful! I just have a quick question here. When i type in the command dfuller *variable*, drift lags(1) regress, the absolute value of the test statistic is greater the the absolute value of the 5% critical value. Does this mean I can say that my time series is stationary?
correct, you are able to reject the null of a unit root, or non-stationarity.
@@mikejonaseconometrics1886 Thank you so much!
Hello. Good video. Thanks for sharing. I have a question. You used the dependent variable all the time right? and not the independent variable?
You should make sure that all variables within your regression model are stationary.
@@mikejonaseconometrics1886 So when we get the first difference of a nonstationary series we generally obtain a stationary one. Now we should use the stationary series which is I(1) and forget about the original nonstationary series. Is this the case?
@@nuricolakoglu4552 Yes, use the difference of the I(1) series in regression models.
@@mikejonaseconometrics1886 Hi. Have ARDL and NARDL replaced OLS based on accuracy of explaining the data?
@@mikejonaseconometrics1886 How can you do this for a binary variable in a probit model?
Hi Mike,
This was really helpful...Could you do one for panel unit root testing---or if you have already, direct me to it?
I'm not including a lagged DV in my panel analysis, so I'm a bit confused as to whether running a unit root test on my panel is necessary. I'm also really confused about how I'm supposed to select the lag structure. I'm using a dicky-fuller test which won't let me use the AIC or BIC protocol, so I have to select a number of lags---but I don't have a predetermined number of lags that I think should be included.
Thanks for making this video. I probably shouldn't be, but I'm always a little befuddled by the difference between drift and trend, and when to include them in the model...
Am I right in thinking that (assuming a positive trend in our series) both manifest themselves in an upward slope? Given the way that time enters into the difference equation at 4mins44 surely this would lead to some sort of exponential growth in the SP500 series, and I would have assumed that this would manifest itself in the time plot of the differenced SP500 as an upward trend.
Your comments would be greatly appreciated!
thank you so much but how can we do all this with 2 variable and compare them??
tsline sp50% regress