Page 1 of 1
AR and K test
Posted: Wed Mar 16, 2011 5:13 am
by aly
Hello! I need to perform Kleibergen's k test and Anderson Rubin test for weak instruments! Can anybody help me?
Thanks
Re: AR and K test
Posted: Wed Mar 16, 2011 7:55 am
by EViews Gareth
EViews (7) only support the related Cragg-Donald statistic.
Re: AR and K test
Posted: Wed Mar 16, 2011 9:40 am
by startz
Also the Anderson-rubin test is for overidentification, not weak instruments.
AR and K test
Posted: Sun Jun 12, 2011 11:46 pm
by aly
Hello! I am trying to implement the Anderson Rubin test, robust to weak instruments in Eviews 7, but get a lot errors.Has anybody tried to implement it? I am asking because I am afraid I am loosing my time and should maybe try in another program. Thank you a lot!
Re: AR and K test
Posted: Mon Jun 13, 2011 7:48 am
by EViews Gareth
Which errors do you receive?
Re: AR and K test
Posted: Thu Jun 23, 2011 6:18 am
by aly
Hello. I am very sorry .I forgot to check if someone answered my question. I managed to solve the pb, I wrongly defined the dimensions of the matrix. Now I have the following questions and hope they are not too stupid. I estimated the following equation equation eq.1gmm inf-c(1)*inf(1)-c(2)*inf(-1)-c(3)*output @ inf(-2 to -4) r_ulc(-1 to -3)
When I construct the AR statistic is something of the following form: (Z'Z)^(-1)/2Z'YQ^(-1)a/(a'Qa)^(1/2)
where a[c(1) 1]' where c(1) is the coefficient of the gmm estimation. My question is: when I construt the statitics how do I call the estimator c(1) within the program? Because when I run the regression I obtain the output with my estimation but within the program I do not knpw how to call the parameter c(1).
My second quustion is: when I construct the matrix of instruments Z, I cannot manage to fill each column with the series that are vectors, I tried something of the following form Z(:,1)=inf_def(-2),Z(:,2)=inf_def(-3), but get error.
Thank you. Hope I did not bother you too much.
Re: AR and K test
Posted: Thu Jun 23, 2011 7:52 am
by EViews Gareth
where
equation is the name of your equation (eq from your code).
Put the instruments into a group, then use either stom or @convert to convert the group into a matrix.
Re: AR and K test
Posted: Thu Jun 23, 2011 11:53 am
by aly
Thank you. I really appreciate it.
Re: AR and K test
Posted: Thu Jun 23, 2011 12:59 pm
by aly
Naturally, I have another question. I am very sorry to bother again. I posted the statistics of AR test in my previous message.The problem is that my matrix of instruments has lagged values of some vectors so when I construct the matrix Z, I have NA values, so I cannot compute the inverse of Z'*Z. Does anybody know how to deal with this pb? I checked some articles and they also ussed lagged values of instruments, so probably have non avaible values in Z.
Thank you again.