[gov2001-l] help - Gov2001 - lists.gking.harvard.edu

jkatkin＠fas.harvard.edu

20 Apr 20 Apr

11:04 a.m.

New subject: [gov2001-l] log-lik value

Hi everyone, I was wondering if it is possible to extract the log-likelihood value for a model from "z.out." In other words, I am really looking for optim.out$value, but would like to avoid having to program the log-likelihood functions myself. Thanks! Jen

Reply

lbarnes＠fas.harvard.edu

21 Apr 21 Apr

5:51 a.m.

New subject: [gov2001-l] Control group

Dear List- Need a clarification: the control groups in ps7 are al those who didn't receive a particular treatment, or only those who received NO treatment of any kind? i.e. are the controls for the mail treatment a: data[which(data$mail ==0),] or b: data[which(data$mail == 0 & data$door == 0 & data$phone == 0),] ? thanks Lucy

Reply

dhopkins＠fas.harvard.edu

6:30 a.m.

New subject: [gov2001-l] MatchIt Arguments

In MatchIt, the "ratio" argument indicates the number of control units to be matched to each treated unit. More generally, the "MatchIt" manual is quite thorough, and provides a good review of how to implement a wide variety of matching estimators. We should have mentioned it in section, and perhaps even assigned it--it will be helpful for the problem set. It is available here: http://gking.harvard.edu/matchit/docs/matchit.pdf Best, Dan ---- Ph.D. Student Department of Government Harvard University Tutor, Currier House dhopkins(a)fas.harvard.edu http://www.danhopkins.org On Fri, 21 Apr 2006, Rebecca Marie Nelson wrote:

...

Yesterday in section we talked about some different options we might select when matching, particularly with vs without replacement and 1 to 1 vs 1 to M matching. How do we control these options in R? I looked at the Zelig guide online and help(matchit), and see that replace=TRUE is an argument - does this control replacement? Any ideas on what argument controls 1 to 1 vs 1 to M matching? I did not see an obvious candidate. Thanks for your help, Becky _______________________________________________ gov2001-l mailing list gov2001-l(a)lists.fas.harvard.edu http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l

Reply

dhopkins＠fas.harvard.edu

6:53 a.m.

New subject: [gov2001-l] Making a dummy variable from continuous variables

In some cases, the cut() variable can be useful. Here, though, I would do something like that following: new.indicator <-1*(abs(var1-var)>=.1) The statement without the 1* provides a logical vector, and the 1* converts it to a 0-1 numeric vector. Best, Dan ---- Ph.D. Student Department of Government Harvard University Tutor, Currier House dhopkins(a)fas.harvard.edu http://www.danhopkins.org On Fri, 21 Apr 2006, Suzanna Chapman wrote:

...

Basic question - I want to make a dummy variable from a continuous variable such that it is attached to my dataframe and can be used in further analysis... but I've only done this before with categorical variables where I could input something like: dummy<-(var1==1)|(var1==2)|(var1==3) dummy[(var1==1)]<-1 dummy[(var1==2)|(var1==3)]<-0 But for a continuous var1, I don't know how to do it. What I want to do is to create a dummy variable that codes 1 when the difference of 2 continuous variables in my data set is greater than or equal to .1. (I thought this would require 2 steps - first making a variable that is the differnce of the two columns, and then making a dummy from that...) difference<- abs(var1-var2) dummy<-as.numeric(div>=.1) But this doesn't create a variable that I can use in regression, etc with my other variables... I get errors that variable lengths differ - even when I try cleaning the data first, etc - so I assume it is because it needs to be attached to the dataset. Any help? On Fri, 21 Apr 2006, Dan Hopkins wrote:

In MatchIt, the "ratio" argument indicates the number of control units to be matched to each treated unit. More generally, the "MatchIt" manual is quite thorough, and provides a good review of how to implement a wide variety of matching estimators. We should have mentioned it in section, and perhaps even assigned it--it will be helpful for the problem set. It is available here: http://gking.harvard.edu/matchit/docs/matchit.pdf Best, Dan ---- Ph.D. Student Department of Government Harvard University Tutor, Currier House dhopkins(a)fas.harvard.edu http://www.danhopkins.org On Fri, 21 Apr 2006, Rebecca Marie Nelson wrote:

Yesterday in section we talked about some different options we might select when matching, particularly with vs without replacement and 1 to 1 vs 1 to M matching. How do we control these options in R? I looked at the Zelig guide online and help(matchit), and see that replace=TRUE is an argument - does this control replacement? Any ideas on what argument controls 1 to 1 vs 1 to M matching? I did not see an obvious candidate. Thanks for your help, Becky _______________________________________________ gov2001-l mailing list gov2001-l(a)lists.fas.harvard.edu http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l

_______________________________________________ gov2001-l mailing list gov2001-l(a)lists.fas.harvard.edu http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l

Reply

dhopkins＠fas.harvard.edu

6:27 a.m.

New subject: [gov2001-l] Control Group and 1(a)

Hi Lucy, Good question--we want you to use those who received no treatment of any kind, to avoid having to worry about confounding alternate treatments. Also, on question 1(a), there has been some confusion about how to estimate the treatment effect, since the model has no covariates but the instructions for estimating the treatment effect mention covariates. Just to be clear, you don't need to worry about covariates in part 1(a)--but we wanted to provide instructions for how to estimate effects that would work later in the problem set, when you will have covariates. Best, Dan ---- Ph.D. Student Department of Government Harvard University Tutor, Currier House dhopkins(a)fas.harvard.edu http://www.danhopkins.org On Fri, 21 Apr 2006, Lucy Barnes wrote:

...

Dear List- Need a clarification: the control groups in ps7 are al those who didn't receive a particular treatment, or only those who received NO treatment of any kind? i.e. are the controls for the mail treatment a: data[which(data$mail ==0),] or b: data[which(data$mail == 0 & data$door == 0 & data$phone == 0),] ? thanks Lucy _______________________________________________ gov2001-l mailing list gov2001-l(a)lists.fas.harvard.edu http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l

Reply