DATA Step, Macro, Functions and more

Descriptive Statistics for Longitudinal (long) Data

Accepted Solution Solved
Reply
Occasional Contributor
Posts: 9
Accepted Solution

Descriptive Statistics for Longitudinal (long) Data

Hi everyone,

 

I have a data set in long format (varied number of observations per subject) with 5 outcome variables. However, not all outcome variables are measures in each observations. I want to get means for each outcome variable the first time it is observed. 

 

Therefore, using first.id does not work (the first observation per subject doesn't necessarily have the first measure of that outcome).

 

Any ideas? 

 

doing

 

IF first.id AND first.outcome then output 

 

does not work either. 


Accepted Solutions
Solution
‎11-16-2017 12:34 PM
Super User
Posts: 23,237

Re: Descriptives Statistics on Longitudinal (Long) Data


All Replies
Occasional Contributor
Posts: 9

Descriptives Statistics on Longitudinal (Long) Data

Hi everyone,

 

I have a data set in long format (varied number of observations per subject) with 5 outcome variables. However, not all outcome variables are measures in each observations. I want to get means for each outcome variable the first time it is observed across subjects. 

 

Therefore, using first.id does not work (the first observation per subject doesn't necessarily have the first measure of that outcome).

 

Any ideas? 

 

doing

 

IF first.id AND first.outcome then output 

 

does not work either. 

Super User
Posts: 23,237

Re: Descriptives Statistics on Longitudinal (Long) Data

Relatively trivial to accomplish this, here are the steps.

 

1. Create flags for each measure, set to 0

2. If this is the first non missing and flag is not set, store it to a new variable and set flag

3. At first iteration of each group set flags and values back to missing. 

 

If you need further help, provide data in a data step so that we can work with data. It doesn't need to exactly your data but needs to reflect what you have. 

 


LGrau wrote:

Hi everyone,

 

I have a data set in long format (varied number of observations per subject) with 5 outcome variables. However, not all outcome variables are measures in each observations. I want to get means for each outcome variable the first time it is observed across subjects. 

 

Therefore, using first.id does not work (the first observation per subject doesn't necessarily have the first measure of that outcome).

 

Any ideas? 

 

doing

 

IF first.id AND first.outcome then output 

 

does not work either. 


 

Occasional Contributor
Posts: 9

Re: Descriptives Statistics on Longitudinal (Long) Data

I'm sorry, I'm not sure what you mean by flags. Here's a sample data set. As you can see, some outcomes are missing until the third observation on the same person. I would need to find the means each outcome the first time it is recorded in each individual.

 

DATA WORK.SOURCE;
INFILE DATALINES DELIMITER= ',' DSD MISSOVER;
INPUT
SubjID
TxGrp :$5.
SexCd :$1.
EthCd
Outcome1
Outcome2
Outcome3;
DATALINES;
101,Arm A,F,1,45,,
101,Arm A,F,1,,3,2
101,Arm A,F,1,43,4,
102,Arm C,M,1,,5,,
102,Arm C,M,1,,4,3
102,Arm B,M,2,32,2,5
;

Solution
‎11-16-2017 12:34 PM
Super User
Posts: 23,237

Re: Descriptives Statistics on Longitudinal (Long) Data

delete_random.JPG

☑ This topic is solved.

Need further help from the community? Please ask a new question.

Discussion stats
  • 4 replies
  • 177 views
  • 0 likes
  • 2 in conversation