Hi all, I know similar questions have been asked about this problem before (see for example
http://www.stata.com/statalist/archi.../msg00531.html). However, my question is not exactly the same.
Without further ado: I have time series cross sectional data and am using psmatch2 with the propensity score option and then with the Mahal option. I'm also familiar with tseffects and nnmatch so if a solution exists with those commands for what I'd like to do, please direct me in that direction.
I have an outcome variable that varies across time and a treatment variable that varies across time, and as the unit 'dyad year' aka say US-Mexico 2000. Assume a dataset that runs from 1980-2010
Assume that the treatment variable is 0 for the US Mexico case until 2005 but then is 1 for years 2005-2010.
with statistical matching, US-Mexico 2005 (treated) can be matched against its own prior values when it has not yet received the treatment (US-Mexico 2000) for example. This is fine and as it should be
However, as a robustness check, I'd also like to ensure that treated cases are matched against other cases that have never received treatment in the time span. So using the US, Mexico example, its treated years of 2005-2010 would only be matched against say US-Canada that is 0 for the entire time span of my data (never received treatment).
Essentially, I'm asking if I could set the control group to 'never treated' cases.
I have toyed around with various roundabout ways for this including exact match option with nnmatch, but none attain what I want it to do. Basically, I'd like to ensure that the control cases are 'never treated' cases but the treatment variable is still time variant.
Is there a way I could do this?or is CEM the way to go ? Thanks much in advance!!
http://www.stata.com/statalist/archi.../msg00531.html). However, my question is not exactly the same.
Without further ado: I have time series cross sectional data and am using psmatch2 with the propensity score option and then with the Mahal option. I'm also familiar with tseffects and nnmatch so if a solution exists with those commands for what I'd like to do, please direct me in that direction.
I have an outcome variable that varies across time and a treatment variable that varies across time, and as the unit 'dyad year' aka say US-Mexico 2000. Assume a dataset that runs from 1980-2010
Assume that the treatment variable is 0 for the US Mexico case until 2005 but then is 1 for years 2005-2010.
with statistical matching, US-Mexico 2005 (treated) can be matched against its own prior values when it has not yet received the treatment (US-Mexico 2000) for example. This is fine and as it should be
However, as a robustness check, I'd also like to ensure that treated cases are matched against other cases that have never received treatment in the time span. So using the US, Mexico example, its treated years of 2005-2010 would only be matched against say US-Canada that is 0 for the entire time span of my data (never received treatment).
Essentially, I'm asking if I could set the control group to 'never treated' cases.
I have toyed around with various roundabout ways for this including exact match option with nnmatch, but none attain what I want it to do. Basically, I'd like to ensure that the control cases are 'never treated' cases but the treatment variable is still time variant.
Is there a way I could do this?or is CEM the way to go ? Thanks much in advance!!
