Hi,
Does anyone know how to extract or calculate the standard errors for random
effects for a mixed-effects linear regression?
The lmer function from lme4 prints out standard errors for the fixed effects
but not for the random effects. I think zelig calls upon lmer, so presumably
won't produce the standard errors I'm looking for. Googling, this seems to
be an issue for several people however I haven't been able to find a
satisfactory answer.
Douglas Bates, who created lmer, says the decision not to report standard
errors for random effects was intentional because he believes they are
misleading. That may be but I wish to calculate them for replication
purposes!
A second question. In zelig, it is possible to run mixed effects logit,
probit and poisson models. I am interested in the possibility of running a
mixed effects oprobit model. How would you go about this, is it even
possible?
Cheerio, Nicola
Hey folks,
I see some of you have already started uploading your data to the Gov2001
Dataverse. This is totally great, but, just in case there's confusion, you
have until the date your turn in your replication to do this.
Maya
Hey folks,
I've set up a Dataverse for Gov2001 final paper replication archives; I've
linked to it from the course webpage, but you can also find it here
http://dvn.iq.harvard.edu/dvn/dv/gov2001
You should all be able to become contributors by starting an account, but if
you run into problems let me know. (This is the first time I'm doing this as
well!) More info on how to become a Dataverse contributor can be found here
http://thedata.org/guides/contributors/contributehttp://thedata.org/guides/contributors/studies
and so on
cool!
Maya
When you get an error message after summarizing a regression model that says:
Coefficients: (14 not defined because of singularities)
What does that mean, exactly?
thanks!!
--
ERIC LIN
Technology and Operations Management
Harvard Business School
Boston, MA 02163
elin at hbs.edu
Hi all,
I was wondering if someone with experience in imputing could help me answer
a question. So I'm trying to impute a TSCS dataset that codes country-years
from 1960-2002. However, I realize that some countries (namely countries
formed from the former Soviet Union and former Yugoslavia) cannot possibly
have information before 1989-ish because they simply didn't exist back then.
Clearly I can't impute the data because in lecture we were told we can't
extrapolate values when there is no underlying answer. Should I just reduce
the focus of the study to 1989 on, or is there some better way to account
for the missingness of these observations?
-Ashley Anderson
Hi all. I am having a problem with using a multicategory treatment
variable in CEM.
I have a survey dataset with mostly dichotomous variables. There are a
few continuous variables (such as age), which I have coarsened a little
bit. The treatment variable (EDUCATION) has three categories. When I
try to match with CEM, I get this error:
Error in imbalance(mat$groups, data = L1data, drop = mat$drop, breaks =
L1.breaks, :
more than 2 groups
The documentation for CEM says that there should be no problem with
multicategory treatment variables, so I am wondering if the dichotomous
control variables, in combination with the mulitcategory treatment
variable, is screwing things up. Or if there is another problem
entirely. When I use a dichotomous treatment variable, I get no errors.
code:
mat3 <- cem(treatment = "EDUCATION", data=cdf.data1.men, drop = c(...),
cutpoints = list(age2 = agecut, INF_MORT = infmcut,
pctvillaccessfootboat = pctvcut), verbose=3)
thanks!
Amanda
Does anyone have any idea where I can find technical information on the models Stata uses? The help files just tell you how to use the command, but they don't tell you anything about the assumptions or the actual model employed.
Thanks,
Nick
Hey folks,
If you're planning on coming to the end-of-the-year Gov 2001 party at Gary's
house (5/8 at noon), please RSVP to the little survey I've set up on the
course website! Do your best to RSVP by lecture next Monday.
thanks!
Maya
I am running the gee package to to logit regressions with clustered standard errors. I can one model to run correctly. There are other models that are run on subsets of the the years in the data. When I run it on subsets, it returns the error rank-deficient model matrix. This seems strange that we'd have rank defficiency just by dropping some observations at the beginning/end of our dataset based on the year. Any help/insight would be greatly appreciated!!
EXL
--
ERIC LIN
Technology and Operations Management
Harvard Business School
Boston, MA 02163
elin at hbs.edu
Hi all,
My qestion is about dependence across observation. In our replication assignment, the authors disaggregated a dependent variable into several dependent variables. That type of coding does make sense in their case. Furthermore, these newly formed dependent variables are mutually correlated. For that reason the authors accounted for the correlations among the newly formed dependent variables by using clustered standard errors. However, the authors did not account for those correlations among observed dependent variables in their model (probit). If that was not the typical way to apply probit model to data analysis, I would consider that an inappropriate use of MLE. Concerning possible solutions, we could account for the correlation among observed dependent variables in different. Two solutions that come to mind are group correlated binary variables with an extended beta-binomial distribution, or ordered probit. Is there anything more appropriate than that? I will appreciate your help.
Nino Malekovic
MPA Candidate, Class 2011
Harvard Kennedy School
________________________________________
From: gov2001-l-bounces at lists.fas.harvard.edu [gov2001-l-bounces at lists.fas.harvard.edu] On Behalf Of gov2001-l-request at lists.fas.harvard.edu [gov2001-l-request at lists.fas.harvard.edu]
Sent: Friday, April 16, 2010 12:00 PM
To: gov2001-l at lists.fas.harvard.edu
Subject: gov2001-l Digest, Vol 57, Issue 22
Send gov2001-l mailing list submissions to
gov2001-l at lists.fas.harvard.edu
To subscribe or unsubscribe via the World Wide Web, visit
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
or, via email, send a message with subject or body 'help' to
gov2001-l-request at lists.fas.harvard.edu
You can reach the person managing the list at
gov2001-l-owner at lists.fas.harvard.edu
When replying, please edit your Subject line so it is more specific
than "Re: Contents of gov2001-l digest..."
Today's Topics:
1. Downloading R on Ubuntu (Shelby Grossman)
2. Re: Downloading R on Ubuntu (Jason Ketola)
3. no problem set this week (Maya Sen)
----------------------------------------------------------------------
Message: 1
Date: Thu, 15 Apr 2010 16:21:00 -0400
From: Shelby Grossman <shelbygrossman at gmail.com>
Subject: [gov2001] Downloading R on Ubuntu
To: "Class List for Gov 2001/E-2001" <gov2001-l at lists.fas.harvard.edu>
Message-ID:
<y2n6dfd2b5a1004151321l359b3f77y29768c65df538f21 at mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
Dear all,
Does anyone have experience downloading R on Ubuntu? I'm having
trouble doing this, even after working through the instructions here:
http://cran.r-project.org/bin/linux/ubuntu/
Many thanks,
Shelby
------------------------------
Message: 2
Date: Thu, 15 Apr 2010 16:39:07 -0400
From: Jason Ketola <jasonketola at gmail.com>
Subject: Re: [gov2001] Downloading R on Ubuntu
To: Class List for Gov 2001/E-2001 <gov2001-l at lists.fas.harvard.edu>
Message-ID: <4BC7796B.1010206 at gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
Hi Shelby,
I run R on Ubuntu.
This is the relevant line from my /etc/apt/sources.list file.
"deb http://software.rc.fas.harvard.edu/mirrors/R/bin/linux/ubuntu karmic/"
Depending on your version, you may need to change the "karmic".
Once you add that line. Run "sudo apt-get update" or run the synaptic
package manager. Then you can install r-base.
J
Shelby Grossman wrote:
> Dear all,
>
> Does anyone have experience downloading R on Ubuntu? I'm having
> trouble doing this, even after working through the instructions here:
> http://cran.r-project.org/bin/linux/ubuntu/
>
> Many thanks,
>
> Shelby
> _______________________________________________
> gov2001-l mailing list
> gov2001-l at lists.fas.harvard.edu
> http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
>
------------------------------
Message: 3
Date: Thu, 15 Apr 2010 20:17:55 -0400
From: Maya Sen <msen at fas.harvard.edu>
Subject: [gov2001] no problem set this week
To: "Class List for Gov 2001/E-2001" <gov2001-l at lists.fas.harvard.edu>
Message-ID:
<k2j16e0be401004151717if6126436vcec7c817a625eb83 at mail.gmail.com>
Content-Type: text/plain; charset="iso-8859-1"
Hi folks,
We made this announcement in section, but in case you weren't around --
there will NOT be a problem set going out this evening or anything due next
Thursday. Instead, we're adding a small assignment onto your replication by
asking you to create a replication data set in the IQSS dataverse (and then
to include a citation to this replication data set in your paper).
We'll be discussing this in lecture on Monday.
best,
Maya