I am trying to match two groups of treatments using Kernal and the nearest neighbor propensity score method . I used the following command in STATA.
psmatch2 RX_cat AGE ERStatus_cat, kernel k(biweight)
psmatch2 RX_cat AGE ERStatus_cat, nn(5)
Where RX_cat stand for treatments, and ERStatus stand for estrogen receptors.
This command gave me the propensity score for each treatment . However, I could not separate the new matched group in a separate variable so I can analyse them separately,i.e. identifying the matched pairs with specific ID.Therefore my question is what the command the I can use to create another column or variable for the matched pairs after assigning a propensity score for them
Rimal,
I use the following code regularly. It creates a set for each matched pair (or matched set). It makes it really easy to then see the data better, and allows you to use clustered standard errors (use vce(cluster "set").
psmatch2 xvar, pscore(pvar) outcome(yvar) caliper(.001) noreplace neighbor(1)
gen pair = _id if _treated==0
replace pair = _n1 if _treated==1
bysort pair: egen paircount = count(pair)
drop if paircount !=2
I hope this helps
Ariel
Here's from stata help file on psmatch2:
In the case of one-to-one and nearest-neighbors matching, for
every treatment observation, it stores the observation number of the
k-th matched control observation. Do not forget to sort by _id if you
want to use the observation number (id) of for example the 1st
nearest neighbor as in
sort _id
g x_of_match = x[_n1]
HTH
Many thanks Prabin:
I tried this before but I did not give me a column for id number
I would say before using psmatch2, run the pscore command to generate the scores first (be sure to include the common support option). Once that's done, you should be able to separate out those obs that are matched
Rimal,
I use the following code regularly. It creates a set for each matched pair (or matched set). It makes it really easy to then see the data better, and allows you to use clustered standard errors (use vce(cluster "set").
psmatch2 xvar, pscore(pvar) outcome(yvar) caliper(.001) noreplace neighbor(1)
gen pair = _id if _treated==0
replace pair = _n1 if _treated==1
bysort pair: egen paircount = count(pair)
drop if paircount !=2
I hope this helps
Ariel
@Ariel Linden : this is a great tool thank you.
For dichotomous outcomes, would you suggest plain logistic regression with "if _weight==1" after psmatch2, or conditional logistic regression using group(pair) created by your code when we want to obtain the odds ratio for treatment effect? Also, shall we use pweights in this final regression if available?
When should I use neighbour(5) instead of neighbour(1) ? I have found that t-stat improves if I use first one. Please explain. Precisely, I have more observation in treated group compared to control group. What integer should I choose for nearest neighbour match?
Dear Poulami,
Kindly, for neighbour(5) instead of neighbour(1), please use the following commands:
psmatch2 xvar, neighbor(4)
gen pair1 = _id if _treated==0
replace pair1 = _n1 if _treated==1
gen pair2 = _id if _treated==0
replace pair2 = _n2 if _treated==1
gen pair3 = _id if _treated==0
replace pair3 = _n3 if _treated==1
gen pair4 = _id if _treated==0
replace pair4 = _n4 if _treated==1
bysort pair1: egen paircount1 = count(pair1)
bysort pair2: egen paircount2 = count(pair2)
bysort pair3: egen paircount3 = count(pair3)
bysort pair4: egen paircount4 = count(pair4)
egen byte paircount = anycount(paircount1 paircount2 paircount3 paircount4), values(2)
drop if paircount==0
Thanks a lot, Shahab. As I understand pair1, pqir2, pair3...are required to drop observations which are out of common support zone. I used one line syntax for this:
psmatch2 Treatment Xs, outcome(outcomeVar) neighbor(5) common ate
drop if _support == 0
will it give different results from yours?
To identify the matched group in the propensity score, I highly recommend following the steps in this link by SSCC Knowledge Base, University of Wisconsin - Madison
https://www.ssc.wisc.edu/sscc/pubs/stata_psmatch.htm
- Badges
- Science topic