tag:blogger.com,1999:blog-2198942534740642384.comments2015-07-05T01:39:18.240-07:00Econometrics Beat: Dave Giles' BlogDave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comBlogger2807125tag:blogger.com,1999:blog-2198942534740642384.post-75327574771820684772015-07-04T03:02:30.045-07:002015-07-04T03:02:30.045-07:00Try this, from the estimated model window, choose ...Try this, from the estimated model window, choose View-->stability diagnostic--> recursive estimationYasmine Rashedhttp://www.blogger.com/profile/17585011442961077074noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-38181369263818108952015-07-03T06:12:04.868-07:002015-07-03T06:12:04.868-07:00The point about the bounds testing is that you may...The point about the bounds testing is that you may have a mixture of I(0) and I(1) variables, but not know for sure - unit root tests have low power. If you do have some I(1) variables, they may or may not be cointegrated - again, you may not know for sure. This ARDL framework enables you to test for a long-run relationship without having all of this information. If you were to do as you you suggest, that would be fine as long as the data were not cointegrated. However, you need the ECT if the I(1) happen to be cointegrated - and you typically don't know for sure. Including the ECT when it's really not needed will not distort the results. However, omitting it when it IS needed leads to a fundamental mis-specification.Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-40797374501442958622015-07-03T03:42:57.667-07:002015-07-03T03:42:57.667-07:00Dear Prof, im verry confused ,,,, if my independan...Dear Prof, im verry confused ,,,, if my independant variables are all in first diffrence and my dependant var is in level , i can make my dependant var also in first diffrence and do a normal ARDL without ECM? is it correct?Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-50549092673494809212015-07-02T14:13:56.957-07:002015-07-02T14:13:56.957-07:00Kerry - my apologies for being so slow in respondi...Kerry - my apologies for being so slow in responding. The use of AIC/SIC for lag length selection should take place after you have dealt with any autocorrelation in the residuals. The reason is this - the AIC and SIC are based on the log-likelihood function, which in turn assumes independence of the observations. This assumption is violated if you have autocorrelation. I don't have a general resource to refer you to. Sorry!Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-75211338037588080472015-07-02T14:05:09.295-07:002015-07-02T14:05:09.295-07:00Please see my reply to "Anonymous" immed...Please see my reply to "Anonymous" immediately below. Given what you've tried already, I'd suspect omitted regressors.Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-6883716272864598472015-07-02T13:51:18.422-07:002015-07-02T13:51:18.422-07:00I've put an example of one of my EViews workfi...I've put an example of one of my EViews workfiles here:<br /><br />http://web.uvic.ca/~dgiles/downloads/VICTORIA_GASOLINE1.WF1Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-13621373688748881092015-07-02T13:35:32.329-07:002015-07-02T13:35:32.329-07:00Yasmine - log(x), I believe.Yasmine - log(x), I believe.Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-66514874144990648732015-07-02T13:15:20.898-07:002015-07-02T13:15:20.898-07:00Thank you - yes it should be, and I have now corre...Thank you - yes it should be, and I have now corrected it.Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-32177864979292174172015-07-02T13:08:50.456-07:002015-07-02T13:08:50.456-07:00Yes you can.Yes you can.Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-83351402931610046622015-07-02T13:02:08.479-07:002015-07-02T13:02:08.479-07:00Also, see http://davegiles.blogspot.ca/2013/07/al...Also, see http://davegiles.blogspot.ca/2013/07/allocation-models-with-bounded.html<br />Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-50551325889319759242015-07-02T13:01:07.007-07:002015-07-02T13:01:07.007-07:00First, suppose that you were using OLS, rather tha...First, suppose that you were using OLS, rather than ANN. In this case you would have an example of what's called an "allocation model". If each equation in the system includes an intercept, then the sum of the dependent variables (the "shares") would equal one. It's easy to show that if you take such a system and estimate each equation by OLS, the PREDICTED values of the dependent variables will all lie between zero and one, and their sum will equal one. This situation arises, for example, with systems of Engel curves. I am not sure how to get this to carry over to ANN. I guess you could model and predict all except for one of the shares. The the forecast for the remaining one would be one minus the sum of the other predictions. I bet the results depend on which share you omit, though - unlike the case with OLS (MLE) where the results are invariant to the one you drop.Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-17728052499283897982015-07-02T10:55:35.432-07:002015-07-02T10:55:35.432-07:00Very useful!Very useful!Unknownhttp://www.blogger.com/profile/12107442801841553114noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-99903233886734612015-07-01T20:59:30.881-07:002015-07-01T20:59:30.881-07:00I am not sure about the derivation of the variance...I am not sure about the derivation of the variance: shouldn't it be (2(k-1)(n-k) ) / ( (n-1)^2(n+1) ) . Thank you Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-24825598565439308422015-07-01T09:41:52.010-07:002015-07-01T09:41:52.010-07:00Yes, you can. Once you have more than 2 variables,...Yes, you can. Once you have more than 2 variables, you have to think about the possibility of "indirect causality".Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-66670254113162954082015-07-01T09:23:47.508-07:002015-07-01T09:23:47.508-07:00I have a question Prof. Giles. I want to test for ...I have a question Prof. Giles. I want to test for Causality using the TY procedure. I have three variables, GDP, Energy Consumption, and CO2 emissions. I was wondering if I could perform this procedure with the three variables and all their possible combination? Cheers, Braulio<br />brauliohttp://www.blogger.com/profile/09142703315035505040noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-22007443927399184802015-07-01T08:30:43.118-07:002015-07-01T08:30:43.118-07:00Yes.Yes.Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-16783082174853313472015-07-01T05:43:12.934-07:002015-07-01T05:43:12.934-07:00Sir can ARDL model be used of all the variables ar...Sir can ARDL model be used of all the variables are I(1) and none is I(0) or I(2) ?Sidrahttp://www.blogger.com/profile/10840247767945130929noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-29991558441972463042015-07-01T04:08:48.795-07:002015-07-01T04:08:48.795-07:00 Dear Sir, I want to know that if we can apply ARD... Dear Sir, I want to know that if we can apply ARDL model in case our dependent variable is stationary and the independent variables are a mix of stationary and non-stationary(integrated at order 1) variables.Mobeen Ur Rehmanhttp://www.blogger.com/profile/16508189296380870891noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-14825696311399650412015-06-30T16:20:18.387-07:002015-06-30T16:20:18.387-07:00Yasmin - There's no "formal" minimum...Yasmin - There's no "formal" minimum. However, keep in mind:<br /><br />1. The testing procedure is based on VAR models - depending on your data frequency, these may require several lags, and this in turn will affect the minimum sample size you can get away with.<br /><br />2. The Johansen procedure is Likelihood-based. Accordingly, good asymptotic properties are assured, but the small-sample power can be low.<br /><br />3. Cointegration is a "long-run" concept. You need a decent "temporal span" in the data to get sensible results. That is, 48 years of annual data may be more relevant than 4 years worth of monthly data. Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-56107563133092944722015-06-30T15:56:46.911-07:002015-06-30T15:56:46.911-07:00Hi Dave,
I have a question related to the topic “M...Hi Dave,<br />I have a question related to the topic “Maximum Likelihood Estimation & Inequality Constraints”, the restrictions on the output of an equation. In some cases the dependent variable can only take certain values (a probability, prices, etc.) therefore some models can be easily adjusted to the data (probit, logit, ANN) or some data transformations can be made.<br />Nevertheless what if one wishes that the output of system of equations satisfy some restriction?<br />I tell you my specify problem, which can shed light on what I say, I have (for a research project) to forecast some shares of sectorial GDP. For this I adjust an Artificial Neural Network for each share, without weights for the activation function, each forecasted share It´s in the range [0,1] but the sum of these is not always equal to one.<br />Best regards,<br />P.S. Thanks for the blog.<br />Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-77177352957858168302015-06-30T15:45:52.059-07:002015-06-30T15:45:52.059-07:00Please see: http://davegiles.blogspot.ca/2015/06...Please see: http://davegiles.blogspot.ca/2015/06/readers-forum-page.htmlDave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-87313423406360392182015-06-30T15:37:30.906-07:002015-06-30T15:37:30.906-07:00Hi Dave,
I have a question related to the topic , ...Hi Dave,<br />I have a question related to the topic , the restrictions on the output of an equation. In some cases the dependent variable can only take certain values (a probability, prices, etc.) therefore some models can be easily adjusted to the data (probit, logit, ANN) or some data transformations can be made.<br /><br />Nevertheless what if one wishes that the output of system of equations satisfy some restriction?<br /><br />I tell you my specify problem, which can shed light on what I say, I have (for a research project) to forecast some shares of sectorial GDP. For this I adjust an Artificial Neural Network for each share, without weights for the activation function, each forecasted share It´s in the range [0,1] but the sum of these is not always equal to one.<br />Best regards,<br />Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-64946924456066685532015-06-30T06:41:19.110-07:002015-06-30T06:41:19.110-07:00Wow, very in depth yet still easy to understand. T...Wow, very in depth yet still easy to understand. Thank you for this.econ manhttp://www.blogger.com/profile/10403319181706638033noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-79413773232054166272015-06-29T12:53:09.593-07:002015-06-29T12:53:09.593-07:00Yes - but you'd have to write an EViews progra...Yes - but you'd have to write an EViews program to do it.Dave Gileshttp://www.blogger.com/profile/05389606956062019445noreply@blogger.comtag:blogger.com,1999:blog-2198942534740642384.post-32349925190793963792015-06-29T12:23:51.613-07:002015-06-29T12:23:51.613-07:00Dear Professor
According to one paper i found Long...Dear Professor<br />According to one paper i found Long-run parameters and standard errors estimaed by ARDL method are biased specially in small sample data. That paper suggested using bias-corrected bootstrap method instead of delta method. Now my question is that how is it possible to do this in eviews 9.0?<br /> Anonymousnoreply@blogger.com