> But do you know how could I change the number of simulations with Zelig?
> I write s.out<-- sim(object, x, x1=NULL, num=c(1000,100), prev=NULL,
> bootstrap=FALSE, bootfn=NULL), it would be increase the iteration number
> (1000), but this doesn�t work with my data set
num = 1000 should do it. see
http://gking.harvard.edu/zelig/docs/commandsRd/sim.pdf for details.
>>> 2- I obtain an output with a different number of treated and
>>> controls in each subclass. In wich way could I match a trated with a
>>> control in each group?
>>>
>>> I have: I
>>> want:
>>> Control
>>> Trated Control Trated
>>> Subclass 1 3000 1600
>>> 1600 1600
>>> Subclass 2 1500 1600
>>> 1600 1600
>>> Subclass 3 1200 1600
>>> 1600 1600
>>> Subclass 4 900
>>> 1600 900 900
>>> Subclass 5 700
>>> 1600 700 700
>>>
>>> I try to do this with subclassification in conjunction with nearest
>>> neighbour, but is not possible
>>>
>>
>> I'm not sure why you would want the same number. What is important in
>> subclassification is to make covariates balanced within each subclass.
>> This objective can be achieved more efficiently if you allow different
>> number of treated and control units in each subclass.
>
>
> Although there are a highest number of control than treated? (see subclass 4
> and 5)
yes, that's fine because the goal is to really improve covariate balance
within each subclass. See our companion paper:
http://imai.princeton.edu/research/preprocess.html
Good luck,
Kosuke
>
>
>> 3-And finally, is possible to export data set after matching in the
>>> way that you can see who is paired with each treated? When I use
>>> write.table, I only obtain the original data set, a column with
>>> distances, a column with weights and the subclass in case of
>>> subclassification method
>>>
>>
>> Have a look at the docs for the function, match.data(), which does
>> this along with other things.
>>
>> Kosuke
>
>
>
> Thanks
>
> ---------------------------------------------------------
>> Kosuke Imai Office: Corwin Hall 041
>> Assistant Professor Phone: 609-258-6601
>> Department of Politics Fax: 973-556-1929
>> Princeton University Email: kimai(a)Princeton.Edu
>> Princeton, NJ 08544-1012 http://imai.princeton.edu
>> ---------------------------------------------------------
>>
>>
>>
>>> I will be very thankful if you could help me
>>>
>>> Sincerly
>>>
>>>
>>> Manuela Exposito Ruiz
>>> Research Support Unit
>>> Medical-Surgical University Hospital
>>> Ja�n Hospital Complex
>>> Spain
>>
>>
>>
>>
>
Dear Listserve
I´m trying to use the MatchIt package with my own files, but I have some
difficulties with subclassification method:
1-After subclassification, I export data set to other statistical software,
and then I estimate the average treatament effect separately in each
subclass. When I try to do this with Zelig package, I get different (but
approximate) results. Could you tell me why? Is cause of number of
simulations in Zelig? In this case, How can I change this option?
2- I obtain an output with a different number of treated and controls in
each subclass. In wich way could I match a trated with a control in each
group?
*I have: I want:*
Control Trated
Control Trated
Subclass 1 3000 1600
1600
1600
Subclass 2 1500 1600
1600
1600
Subclass 3 1200 1600
1600
1600
Subclass 4 900 1600
900
900
Subclass 5 700 1600
700
700
I try to do this with subclassification in conjunction with nearest
neighbour, but is not possible
3-And finally, is possible to export data set after matching in the way that
you can see who is paired with each treated? When I use write.table, I only
obtain the original data set, a column with distances, a column with weights
and the subclass in case of subclassification method
I will be very thankful if you could help me
Sincerly
Manuela Exposito Ruiz
Research Support Unit
Medical-Surgical University Hospital
Jaén Hospital Complex
Spain