Rimal, it should be reported as a column designated: _weight.

Many thanks for your reply Joshua. I do not have this column. The only designated columns that I have got are  myscore, myblock, comsup.The commands that I used are

pscore RX_cat AGE,pscore(myscore) blockid (myblock) comsup

attnd totalcost_alltreatment RX_cat AGE,pscore(myscore) comsup.

How can I get it

Rima, I am not as familiar with pscore. I typically use psmatch2. It is a user generated command, I believe, but can be found with the findit command in Stata. It will generate groups that are easily differentiable and the treatment effect is obvious:

http://www.ssc.wisc.edu/sscc/pubs/stata_psmatch.htm

Best of luck,

Joshua

Dear Joshua

Many thanks for your help.It is highly appreciated.

Rimal

Rimal, you may want to consider using -optmatch2-  (a user-written program. from within Stata, type: findit optmatch2)

What is great about this command is that a variable is automatically generated that will give you the ID of the grouping (or pairing). Thus you don't need to go to the extra steps that psmatch2 requires of you to identify and match up the controls

I hope this helps

Ariel

I forgot to mention that Mark Lunt, who wrote optmatch2, wrote some other propensity score matching commands as well. They all provide that ID variable I mentioned above. I happen to prefer optmatch2 over the other commands.

From within Stata, type:

net from http://personalpages.manchester.ac.uk/staff/mark.lunt

Dear Ariel:

Many thanks for your effort to help me. It is highly appreciated.

The manual seems very useful .I will take your advice and follow the manual recommendations  .Many thanks also for up voting my question.

I did an alternative way to propensity score but I am not sure if it will be sufficient or not to make the two group comparable :

first I separate the patients in the non -surgery group into four age categories i.e. less than 75, 75-85,85-95.

Second, I used the literature to identify the percentages of patients who are high likely to be fit for surgery within each age category. e.g. literature showed that 30% of people between 75-85 are frail therefore 70% are fit to surgery.

Third, I picked up a random sample of 70% without replacement using the sample command in STATA and I discarded the 30% .I did the same for each age category and then I combine the age category together.

Fourth,I used multivariate analysis to adjust for age between both treatment groups.

I am not sure I understand your process? The propensity score is estimated using a logisitic (or probit) regression model, where the binary outcome of the model is whether the individual did or not receive the intervention conditional on covariates. So in your data, the outcome of the logit model should be surgery/no-surgery. From your description, I am not sure how the percentage of fraily elderly is used here? Also, I am not sure how the "random sample of 70%" is used here? Finally, I don't understand what you did with the age categories that you needed to merge them together?

One of the reason for receiving non surgery is due to frailty, therefore, I am trying to exclude the patients who are likely to be frail by using age as surrogate indicator and match the m with patients who underwent surgery . In the literature ,70% of elderly who aged between 75 and 85 are frail. In STATA when you use command 'sample 70' if age between 75-85, STATA will pick up 70% of the population and discard the others.I did the same for each category. Therefore , the final population will be patients who are not frail.

I am not sure if propensity score will fit my purpose because I do not have the exact frailty status or comorbidities  for each patient

Hope this clear

There are two options: First is to simply add

keep if _weight == 1

That should keep only the treated cases.

// or you can apply:

gen pair = _id if _treated==0

replace pair = _n1 if _treated==1

bysort pair: egen paircount = count(pair)

drop if paircount !=2