[R] HELP - Hausman Test + systemfit
Arne Henningsen
arne.henningsen at gmail.com
Wed Jan 6 22:08:42 CET 2016
Dear Kateryna
The hausman() method for objects of class "systemfit" tests
multiple-equation models estimated by 3SLS against multiple-equation
models estimated by 2SLS (see documentation, e.g. [1]). It seems to me
that this is not the type of Hausman test that you want to conduct.
[1] http://www.inside-r.org/packages/cran/systemfit/docs/hausman.systemfit
Best regards,
Arne
On 6 January 2016 at 20:02, David Winsemius <dwinsemius at comcast.net> wrote:
>
>> On Jan 6, 2016, at 9:44 AM, Jeff Newmiller <jdnewmil at dcn.davis.ca.us> wrote:
>>
>> To post appropriately, use less apologizing and more reading of the Posting Guide that is mentioned in the footer of all messages on this list (including this one). Note that HTML does not work well on the list, so be sure to turn it off in your email program (at least when sending emails here).
>>
>> I have never used this function, but the error message is pretty clear. If you read the help for the systemfit function it states "list of formulas", not "list of model objects". The objects returned from lrm are not formulas. Perhaps look at some examples of using that function, such as the ones in the help (type ?systemfit after you have loaded the systemfit package into R).
>
> To go a bit further in explaining what might have better prepared responders to offer a specific and tested reply, please note as you read through the Posting Guide that a data example is strongly suggested. Also suggested is that you post all `library` calls for any packages beyond the base set. The `lrm` function, for instance` is most likely from the 'rms'-package. That raises a further concern about which I have limited experience to offer. The rms-regression functions may or may not have methods defined that allow use of functions from other packages. I see that when 'systemfit' is loaded, the `lrtest`-function from rms is listed as being masked, and I further see that lrtest {lmtest} assumes there will be an `update`-method, but I do not see an `update.lrm` function when I execute:
>
> methods(update)
>
> On the other hand there may be grounds for guarded optimism. After preparing the example on the help page for `lrm` and creating the lrm example 'fit' we see that lrm objects may be constructed so that the can inherit the glm class and therefore use the update.glm method.
>
>> class(fit)
> [1] "lrm" "rms" "glm"
>
> So to proceed, you might make a list of those models and then run `lapply` with the `formula` function to extract the needed formula-objects:
>
> formula(fit)
> #y ~ blood.pressure + sex * (age + rcs(cholesterol, 4))
>
> --
> David.
>
>
>> --
>> Sent from my phone. Please excuse my brevity.
>>
>> On January 6, 2016 5:02:52 AM PST, Kateryna Riabchenko <riabchenko.kateryna at gmail.com> wrote:
>>> Hello, I am not advanced in R (that is why my questions can sound
>>> stupid).
>>> I apologize for that in advance. But that is how my brain works - I
>>> need to
>>> ask questions to understand. I was searching the answer everywhere -
>>> without result. So I am asking You.
>>>
>>> *Given:*
>>>
>>> I have 8 regression Models:
>>> Model1 <- lrm(ACINTENSITY ~ GDPGROWTH, data = ACU)
>>> Model2<- lrm(ACINTENSITY ~ GDPGROWTH + POPULATION, data = ACU)
>>> Model3 <- lrm(ACINTENSITY ~ GDPGROWTH + POPULATION + FEDUNION, data =
>>> ACU)
>>> Model4<- lrm(ACINTENSITY ~ GDPGROWTH + POPULATION + FEDUNION +
>>> CENTRALBANK,
>>> data = ACU)
>>> Model5 <- lrm(ACINTENSITY ~ GDPGROWTH + POPULATION + FEDUNION +
>>> CENTRALBANK
>>> + CPI, data = ACU)
>>> Model6 <- lrm(ACINTENSITY ~ GDPGROWTH + POPULATION + FEDUNION +
>>> CENTRALBANK
>>> + CPI + INTERATE, data = ACU)
>>> Model7<- lrm(ACINTENSITY ~ GDPGROWTH + POPULATION + FEDUNION +
>>> CENTRALBANK
>>> + CPI + INTERATE + UNEMPL, data = ACU)
>>> Model8 <- lrm(ACINTENSITY ~ GDPGROWTH + POPULATION + FEDUNION +
>>> CENTRALBANK
>>> + CPI + INTERATE + UNEMPL + INTERNET, data = ACU)
>>> As you can see, each time a new explanatory variable is added.
>>> I want to perform the Hausman Test
>>> for that I installed systemfit package
>>> My idea was to compare Model1 with Model2, Model2 with model3, ... and
>>> so on
>>>
>>> *I started with this*:
>>> inst <- ~ POPULATION - *this is Question1:* I don't know what I need
>>> to
>>> mention in *"inst"*. I put additional variable in Model2 which does not
>>> exist in Model1, but I ma not sure that is correct! Or maybe I need to
>>> put
>>> all variables, which were used in 2 models. if You can explain - thank
>>> you!
>>> system <- list(Model1, Model2)
>>>
>>> # perform the estimations
>>> fit2sls <- systemfit(system, "2SLS", inst = inst, data = ACU)
>>>
>>> but R responded:
>>>
>>> Error in systemfit(system, "2SLS", inst = inst, data = ACU) :
>>> the list of argument 'formula' must contain only objects of class
>>> 'formula'
>>>
>>>
>>>
>>> Please, help me to understand What I do wrong!
>>> Best,
>>> Kateryna
>>>
>>> [[alternative HTML version deleted]]
>>>
>>> ______________________________________________
>>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>> PLEASE do read the posting guide
>>> http://www.R-project.org/posting-guide.html
>>> and provide commented, minimal, self-contained, reproducible code.
>>
>> [[alternative HTML version deleted]]
>>
>> ______________________________________________
>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>
> David Winsemius
> Alameda, CA, USA
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
--
Arne Henningsen
http://www.arne-henningsen.name
More information about the R-help
mailing list