[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: simple sum() question

From   Shehzad Ali <>
Subject   Re: st: simple sum() question
Date   17 Apr 2009 13:33:33 +0100

Following on after yesterday's discussion, I have a quick follow on question.

Here is a quick summary of what I am doing. Each patient (varname: patient_id) was observed at 4 time points and at each time point we asked about the clinic visits (varname: clinic) in the last 3 months. The dataset is in long form (shown below):

patient_id timepoint clinic 1 1 0
1                 2           1
1                 3           .
2                 1           2
2                 2           0
3                 1           1
3                 2           .

The line below generates a sum of all clinic visits for each patient:

bysort patient_id: egen sum_clinic = sum(clinic)

Now if at one time point, clinic visit is missing (as its seen for patients 1 and 3), then I want stata to return missing value for the sum. The above command returns the total of the non-missing observations, ignoring the missing ones (understandably). But if I tried:

bysort patient_id: egen sum_clinic = sum(clinic) if clinic!=.

then it returns missing value for the sum variable only for the time point which is missing and not for all the time points for that patient. Can anyone please suggest how to resolve this?

Secondly, whats the best way to collapse the dataset to one observation per patient? Once I have the sum_clinic for each patient, it would be easier just to have one observation per patient.

Thank you,

On Apr 15 2009, Martin Weiss wrote:


Those two differ only in case you have missings...

----- Original Message ----- From: "Shehzad Ali" <>
To: <>
Sent: Wednesday, April 15, 2009 7:06 PM
Subject: RE: st: AW: simple sum() question

Thanks, Nick. But I am not trying to count the total number of observations per patient but the total number of visits (varname: clinic) across all time points for each patient (I tried to clearly state it in the first post - sorry if I wasn't clear).

The solution I am now using is:

bysort patient_id: egen sum_clinic = sum(clinic)

Thank you,


On Apr 15 2009, Nick Cox wrote:

Unless there are further complications as yet unrevealed,
bysort id : gen visits = _N
is a direct and simple solution.
If you just wanted to count a subset, then
gen interesting = <binary variable defining interesting> bysort interesting id : gen interesting_visits = _N if interesting
There are -egen- routes as well, but for problems like this going back
to basics is difficult to beat.
See also
SJ-2-1  pr0004  . . . . . . . . . . Speaking Stata:  How to move step
by: step
       . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  N.
J. Cox
       Q1/02   SJ 2(1):86--102                                  (no
       explains the use of the by varlist : construct to tackle
       a variety of problems with group structure, ranging from
       simple calculations for each of several groups to more
       advanced manipulations that use the built-in _n and _N

if a tutorial is needed. That's free on-line at the Stata Journal
Note that even if you did want a collapsed dataset, -contract- rather
than -collapse- is more direct.
Shehzad Ali

Hi Martin and Josiane,

Thank you for your replies. You are right that I am interested in the
total count of visits for each patient and not the running sum.

Sorry, I should have mentioned that patients who had three visits, for instance, have three observations, and those with two visits have two observations. Therefore, the total number of observations for 100
patients is less than 400 (I had made up hypothetical numbers in haste to
simplify the case. Not always a good idea).

With Martin's solution, I will need to have four observations for each patient (sorry this was my fault as I didn't provide the correct information). With Josiane's suggestion, the dataset collapses which is
not what I want.

Can you suggest a modified solution please? Again, sorry for the unclear

email earlier.

On Apr 15 2009, Martin Weiss wrote:

I am betting that you want a count of visits, not a running sum, but correct me if I am wrong...

set obs 400
egen float patient = seq(), from(1) to(400) block(4)
egen float visit = seq(), from(1) to(4) block(1)

//not strictly necessary
xtset patient visit

//less than 4 visits for some
replace visit =. if runiform()<0.05

bys patient: egen overallvisits=count(visit)

l in 1/20, sepby(patient) noo

Shehzad Ali

I have a simple question about summing across observations. I have 100 patients (variable: patient_id) in the dataset, each had clinic visits (variable: clinic) and hospital visits (variable: hospital) recorded at

weeks 4, 8, 12 and 16. The dataset is long and hence I have 400 observations (one observation per patient per time point).

I want to sum the clinic visits for each patient (across all 4 visits)

bearing in mind that some patients had less than 4 visits. So
I want to generate a new variable that will produce the sum of clinic visits for each patient.

*   For searches and help try:

*   For searches and help try:

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2020 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index