tag:blogger.com,1999:blog-2198942534740642384.comments2019-10-14T17:33:39.796-07:00Econometrics Beat: Dave Giles' BlogDave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comBlogger4216125tag:blogger.com,1999:blog-2198942534740642384.post-3339239192552684832019-10-14T07:33:52.120-07:002019-10-14T07:33:52.120-07:00Hi - Check out https://www.stata.com/support/faqs/...Hi - Check out https://www.stata.com/support/faqs/statistics/linear-regression-with-interval-constraints/<br /><br />and https://www.stata.com/statalist/archive/2008-06/msg00217.html<br />Dave Gileshttps://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-58877541880286387062019-10-14T07:23:57.962-07:002019-10-14T07:23:57.962-07:00Thank you for your in-depth explanation. Do you ha...Thank you for your in-depth explanation. Do you have suggestions for how to impose non-positive constraints on coefficients in Stata (using either the nl or cnreg commands)?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-54531798899144391132019-10-13T11:41:23.377-07:002019-10-13T11:41:23.377-07:00Hello
If anyone have codes for FIVAR in Eviews.......Hello<br />If anyone have codes for FIVAR in Eviews.... kindly share...Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-6391254180231282542019-10-13T11:15:51.766-07:002019-10-13T11:15:51.766-07:00hello. anyone here who can share code for *rtadf* ...hello. anyone here who can share code for *rtadf* for Eviews? Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-31719026117966259062019-10-09T13:20:25.899-07:002019-10-09T13:20:25.899-07:00On page 90 of Barreto and Howland (2006), one can...On page 90 of Barreto and Howland (2006), one can read<br /><br />“Finally, note that we employ Excel as a teaching tool and do not recommend its use for serious work in econometrics. Excel can usually (not always!) be relied on to provide correct results for simple analyses such as would be found in the typical term paper in introductory econometrics classes. To be assured of accuracy, however, it is best to use a specialized software package”<br /><br />I take it from this that they agree with the sentiments of this blog. I totally agree<br />Unknownhttps://www.blogger.com/profile/09231128516677196072noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-19800256690636850502019-10-03T07:09:48.372-07:002019-10-03T07:09:48.372-07:00You might find this paper helpful.You might find this paper helpful.Dave Gileshttps://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-91427461324300010372019-10-03T04:05:02.789-07:002019-10-03T04:05:02.789-07:00Dear Prof, could you kindly explain to me how mixe...Dear Prof, could you kindly explain to me how mixed data sample regression (MIDAS) is used as granger causality techniqueUnknownhttps://www.blogger.com/profile/03714019615659133394noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-37833111218924026962019-09-27T06:51:00.742-07:002019-09-27T06:51:00.742-07:00Hi Dr. Giles,
I have what might be a relatively e...Hi Dr. Giles,<br /><br />I have what might be a relatively elementary question, but I've been told two different answers.<br />Normally in an IV situation, the instrument needs to be as good as randomly assigned, conditional on covariates. And you need at least as many instruments as endogenous variables.<br />So I would like to estimate the causal effect of the Alberta economy on the Ontario economy. Obviously I cannot just run an equation of GDP_ON = GDP_AB do to the number of Canada wide shocks that would hit both. So for an instrument that would affect Alberta but not Ontario I was thinking of using refinery shut-downs (such as Fort McMurray 2016).<br />However, the other way that the refinery shut-downs could affect the Ontario economy is through the depreciation of the Canadian dollar.<br />If I were to include the exchange rate as a control (and assuming no other causal channels), do I have an identification strategy? The reason why I'm not 100% is that it seems like I have two endogenous variables and only 1 instrument. Could you provide insight on the matter? Additionally if you see other reasons why this wouldn't work I would be happy to realize that now.<br /><br />Awesome blog! Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-89471095148825690322019-09-26T04:04:54.592-07:002019-09-26T04:04:54.592-07:00It should be a fixed regressor. After all, it'...It should be a fixed regressor. After all, it's just "shifting" the intercept in the model, and the intercept itself is a fixed regressor.Dave Gileshttps://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-69593482516116176732019-09-26T01:14:32.750-07:002019-09-26T01:14:32.750-07:00Dear Professor, can I include dummy variable in AR...Dear Professor, can I include dummy variable in ARDL estimation? Or it should be a fixed regressor?Vickienoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-12012802007661490992019-09-11T08:04:30.524-07:002019-09-11T08:04:30.524-07:00Yes, you should.Yes, you should.Dave Gileshttps://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-48004595197472876232019-09-10T21:38:03.717-07:002019-09-10T21:38:03.717-07:00This article helped me a lot. Thank you! Should we...This article helped me a lot. Thank you! Should we use Finite Population Correction Factor (FPC) when you sample without replacement?Unknownhttps://www.blogger.com/profile/11950817972447602623noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-16237877565901478232019-09-07T04:03:53.217-07:002019-09-07T04:03:53.217-07:00Thank you!Thank you!Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-24588408591588243932019-09-06T20:31:27.544-07:002019-09-06T20:31:27.544-07:00The method presented in "A cheap trick to imp... The method presented in "A cheap trick to improve the power of a conservative hypothesis tests" should be used with caution in practice. It is not universal applicable. They used the "same" in their time series paper several years ago. If it was true, it should be a big result not a simple one. Until now, no one cited them...Unknownhttps://www.blogger.com/profile/11688604215091070232noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-88181639428796202352019-09-06T10:36:21.221-07:002019-09-06T10:36:21.221-07:00Amir - I have no idea why they would differ. Perha...Amir - I have no idea why they would differ. Perhaps you should check with the producers of the 2 packages. Ideally, confidence intervals should be given. You could bootstrap them.Dave Gileshttps://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-50448746859854367952019-09-06T09:58:47.035-07:002019-09-06T09:58:47.035-07:00Prof Giles,
Thank you for introducing me to Gretl!...Prof Giles,<br />Thank you for introducing me to Gretl!<br /><br />I have a question regarding IRF for VECM on Eviews and Gretl.<br />On Eviews the IRF (for two variables) gives a decline in Y before a later increase in response to shocks to X. However in Gretl, Y increases in response to shocks to X.<br />I tried my best to apply the same setting regarding VECM. In your opinion, what's the reason for contradictory results?<br /><br />One more thing, Do you believe it is correct to report VECM IRF results from Eviews (even though they don't show confidence intervals) in a paper?<br /><br />I would appreciate your comment!<br /><br />Amir,Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-56547332666802741602019-08-30T06:11:16.617-07:002019-08-30T06:11:16.617-07:00Often, that will be sufficient, depending on how &...Often, that will be sufficient, depending on how "complicated" your break is.Dave Gileshttps://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-78998560139276013802019-08-29T21:21:39.951-07:002019-08-29T21:21:39.951-07:00Prof Giles, to incorporate structural breaks in co...Prof Giles, to incorporate structural breaks in cointegration and VECM, is it enough to add a dummy variables for a specific year in the exogenous box on VAR, Cointegration, and VECM?Amir https://www.blogger.com/profile/08355544087004838174noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-27050180416566221122019-08-29T04:16:05.039-07:002019-08-29T04:16:05.039-07:00Thank you Prof. Giles for your suggestions!
I ass...Thank you Prof. Giles for your suggestions!<br /><br />I assume my reply haven't gone through. So I'm replying again.<br />___<br /><br />So, are the following formulas correct, professor?<br /><br />1- Y and X cointegrated: ____<br /><br />Estimating regression: y(t) = b(0) + b(1)*x(t) + e(t)<br />=> e(t)= y(t)-b(1)*x(t)-b(0)<br /><br />Estimating VECM: d(y(t)) = c(1)*e(t-1) + c(2)*d(y(t)) + c(3)*d(x(t)) + C(4) + u(t)<br /><br />a)<br />Can I interpret C(3) as 1% increase in the annual changes in X cause c(3)% increase in Y?<br /><br />b)<br />Should I do the estimation manually?<br />Because When I allowed Eviews to estimate a VECM (1 1 lag) the coefficients of the ECT were slightly different, and also there was a trend:<br />D(Y) = C(1)*( Y(-1) - B(1)*PD(-1) - b(2)*@TREND(80) + b(3)) + C(2)*D(Y(-1)) + C(3)*D(X(-1)) + C(4)<br /> <br /><br />2- Y and X not Cointegrated: ____<br /><br />Estimating d(y) = d(x) + c<br /><br />a)<br />According to Jarque-Bera test, the residuals are not normally distributed. Does that affect the interpretation of the coefficient of X?<br /><br />b)<br />Should I instead estimate with VAR and choose the maximum lag suggested?<br /><br /><br />Thank you so much for the effort and time you put into this blog!Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-34793444574040627842019-08-27T03:19:03.089-07:002019-08-27T03:19:03.089-07:00Dear Professor,
When comparing two regressions, ...Dear Professor, <br /><br />When comparing two regressions, how can the Standard Deviation of the dependent variable and residual sum of squares be used. If the values are lower does this mean the dependent variable is more predictable. Thanks in advanceuser413535https://www.blogger.com/profile/01318406039619327046noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-21488477589377030352019-08-26T11:50:39.965-07:002019-08-26T11:50:39.965-07:001. If you have only 2 series, then there can't...1. If you have only 2 series, then there can't be more than one cointegrating relationship. Why not just estimate an OLS regression using the levels of the data to get the long-run equilibrating relationship; and estimate a standard single-equation erro-correction model to get the short-run dynamics?<br />2. In this case regress the first-difference of y on the first-difference of x, using OLS. If needed, add lags of these two differenced series as extra regressors to eliminate any autocorrelation in the errors.Dave Gileshttps://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-52604320252356513722019-08-26T11:47:46.402-07:002019-08-26T11:47:46.402-07:00Dear Prof. Giles,
According to your experience ho...Dear Prof. Giles,<br /><br />According to your experience how to examine the short-run relationship between y and x in the following two scenarios: (both are I(1))<br /><br />1- Only one cointegration exists. Hence a VECM (which subsequnetly will have to be lagg 0 0) would not show the short-run dynamics (of dX(t0-1)). What would you recommend to do in this case?<br /><br />2-No cointegration exists. So what is the best way to examine whether y responses to x in the short run?<br /><br />Statistical package used: Eviews.<br /><br />Thank you!Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-63725397227959440862019-08-24T07:52:16.918-07:002019-08-24T07:52:16.918-07:00You cannot have ANY I(2) variables anywhere in the...You cannot have ANY I(2) variables anywhere in the model.Dave Gileshttps://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-81987560577049518182019-08-24T07:51:29.928-07:002019-08-24T07:51:29.928-07:00There could be several reasons. Your errors may be...There could be several reasons. Your errors may be autocorrelated, for instance, and then the tests won't be appropriate. You may have mis-specified the lag structure (do you have enough lags included?) If you have a fairly small sample then this could be the reason.Dave Gileshttps://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-83059487751326342782019-08-24T07:48:36.351-07:002019-08-24T07:48:36.351-07:00This comment has been removed by the author.Dave Gileshttps://www.blogger.com/profile/05389606956062019445noreply@blogger.com