About luch25

luch25 · ‎01-29-2022

Hello all, I'm looking for some advice on how to go about coding this. I have a dataset with longitudinal measurements of heart rate (HR) and respirations (resp). Each individual (ID) has multiple measurements. The three main things I need to do are: By unique ID, flag measurements that are taken within 5 minutes of each other (flag=1) For measurements that are NOT taken within 5 minutes of another measurement, retain the original value of HR and resp. For measurements that are taken within 5 minutes of each other, apply these rules to produce the final dataset If there are 2 measurements within the 5-minute interval, take the average of the measurements, as well as the datetime value to produce a new value If there are >2 measurements within the 5-minute interval, delete the first value, and take the average fo the remaining measurements. Code for an example dataset are provided below. I've also provided a description of "what's going on" for each individual below the code. data have; input id datetime hr resp; format datetime datetime22.3.; cards; 1 19MAR2009:13:15:00.000 90 16 1 19MAR2009:13:17:00.000 70 18 1 19MAR2009:13:18:00.000 80 17 1 19MAR2009:13.22:00.000 100 22 1 23MAR2009:10:36:00.000 83 21 2 29OCT2009:10:36:00.000 121 13 2 29OCT2009:10:38:00.000 83 14 2 08JUL2011:13:04:00.000 131 18 3 07SEP2016:14:26:59.000 50 12 3 06APR2017:13:39:00.000 76 18 3 08JUL2017:13:04:00.000 98 17 3 08JUL2017:13:08:00.000 96 18 3 24OCT2017:15:11:00.000 80 19 4 18APR2012:08:42:00.000 80 17 4 31OCT2012:10:36:00.000 96 19 4 31OCT2012:10:37:00.000 103 21 4 31OCT2012:10:38:00.000 105 25 4 31OCT2012:10:39:00.000 110 28 5 06JAN2008:05:37:00.000 98 17 5 15FEB2010:10:24:00.000 110 18 ; ID=1 has two "5 minute intervals". The first 3 measurements make up one interval (delete the one at 13:15 and take the average of 13:17 and 13:18. Then the 2nd, 3rd, and 4th measurements make up a second interval (delete the one at 13:17, and take the average of 13:18 and 13:22). ID=2 has one "5 minute interval" with 2 measurements, so take the average of these two ID=3 has two measurements within a "5 minute interval" ID=4 has 4 measurements within a "5 minute interval". Delete the one at 10:36 and take the average of the remaining ones ID=5 does not have any measurements within a 5 minute interval And as you can see, I want to retain measurements even if they are not in the 5 minute interval (flag=0)! Applying these rules, this is what I envision the final dataset looking like: ID Datetime (format=datetime22.3) HR2 Resp2 Flag 1 19MAR2009:13:17:30.000 75 17.5 1 1 19MAR2009:13.19:00.000 83.3 19 1 1 23MAR2009:10:36:00.000 83 21 0 2 29OCT2009:10:37:00.000 102 13.5 1 2 08JUL2011:13:04:00.000 131 18 0 3 07SEP2016:14:26:59.000 50 12 0 3 06APR2017:13:39:00.000 76 18 0 3 08JUL2017:13:06:00.000 97 17.5 1 3 24OCT2017:15:11:00.000 80 19 0 4 18APR2012:08:42:00.000 80 17 0 4 31OCT2012:10:37:30.000 103.5 23.25 1 5 06JAN2008:05:37:00.000 98 17 0 5 15FEB2010:10:24:00.000 110 18 0 Any advice on how to handle this would be hugely appreciated! Thank you in advance!

luch25 · ‎01-13-2022

Thank you for your response! Yes, I think this is working now.

luch25 · ‎01-12-2022

Thank you! Yes, you're correct, that was an error on my part. I'm getting an error message, "Array subscript out of range" for the "dt[i]=date" portion of the code. Any suggestions on how I can troubleshoot this?

luch25 · ‎01-12-2022

2. Yes, the start date is the minimum for date for the id 3. An encounter does not have to be exactly at 2 years. It just needs to be within the 2 year time frame 1. Sorry, I should have included the desired output. In addition to the "flag" and "first_enc_date" columns from your code, I would like time1-timei columns that indicate if, during the 2 year period, the individual had at least one encounter. So for id=1 below, the time-based variables would be created as noted below. 0 means the individual does not have at least one encounter during the timeframe. 1 means the individual has at least one encounter. time1 = 19SEP2005-18SEP2007 = 1 time2 = 19SEP2007-18SEP2009 = 1 time3 = 19SEP2009-18SEP2011 = 0 time4 = 19SEP2011-18SEP2013 = 0 time5 = 19SEP2013-18SEP2015 = 1 time6 = 19SEP2015-18SEP2017 = 1 id date flag first_enc_date time1 time2 time3 time4 time5 time6 1 19SEP2005 . 19SEP2005 1 1 0 0 1 1 1 09NOV2007 . 19SEP2005 1 1 0 0 1 1 1 11JAN2008 1 19SEP2005 1 1 0 0 1 1 1 14FEB2009 1 19SEP2005 1 1 0 0 1 1 1 15JUN2015 . 19SEP2005 1 1 0 0 1 1 1 15JUN2016 1 19SEP2005 1 1 0 0 1 1 2 13DEC2011 1 13DEC2011 1 1 2 15JUN2012 1 13DEC2011 1 1 2 17NOV2012 1 13DEC2011 1 1 2 17SEP2013 1 13DEC2011 1 1 2 20JAN2015 1 13DEC2011 1 1 3 18JAN2003 . 18JAN2003 1 1 3 22MAR2005 . 18JAN2003 1 1 Please note that I modified the original data for studyid=1. Here's the code for the output that I listed above: data have; input id date : date9.; format date date9.; cards; 1 19SEP2005 1 09NOV2007 1 11JAN2008 1 14FEB2009 1 15JUN2015 1 15JUN2016 2 13DEC2011 2 15JUN2012 2 17NOV2012 2 17SEP2013 2 20JAN2015 3 18JAN2003 3 22MAR2005 ; Thank you for your time and help @r_behata !

luch25 · ‎01-11-2022

Hello all! I have a longitudinal dataset with repeated measures (healthcare encounter dates) for individuals. The individuals all have different "start dates" (ie, first healthcare encounter date), and they can have differing number of total encounters. I need to a criteria that individuals in my study have at least one healthcare encounter every 2 years between their start (first healthcare encounter date) and end date (last healthcare encounter date). I was thinking about doing this by creating arrays that are 2 years long based on each person's start date, and then "flagging" if an individual has at least one encounter within each array. Any advice/resources on how to go about this? Thank you in advance! data have; input id date; cards; 1 19SEP2005 1 09NOV2007 1 11JAN2008 1 14FEB2009 1 15JUN2011 2 13DEC2011 2 15JUN2012 2 17NOV2012 2 17SEP2013 2 20JAN2015 3 18JAN2003 3 22MAR2005 ;

luch25 · ‎09-18-2020

Thank you! This works perfectly. Should have thought of proc summary

luch25 · ‎09-18-2020

Hi, I need to be able to count the number of individuals within a state that have a certain option. I know how to do this in proc sql for a single option, but not sure how to construct this if I want counts by states for multiple (i.e., 5) options. I could merge 5 separate proc sql, but I'm wondering if there's a way to do this in a single step. Here's sample SAS code for the dataset: data have; input id state option1 option2 option3 option4 option5; cards; 1 1 1 0 1 1 1 2 2 0 0 1 1 1 3 3 0 1 0 1 0 4 1 1 1 0 1 1 5 2 0 1 1 0 1 6 3 1 1 1 1 1 7 1 1 0 0 0 1 8 2 0 1 0 1 0 9 3 1 1 1 1 1 ; And this is the dataset that I'm trying to obtain. Note that the counts of individuals are by each state. state option1 option2 option3 option4 option5 1 3 1 1 2 3 2 0 2 2 2 2 3 2 3 2 3 2 Thanks for any advice that you have!

luch25 · ‎08-31-2020

Yes, thank you!!

luch25 · ‎08-31-2020

Hello, I need to concatenate 100 columns of numeric values. An small example dataset is provided below. I want the concatenated variable to be separated by a space, so the variable reads like this: "41 42 43 44 45 46 47"; "61 62 63 64". Note that some rows have missing values (they don't have a value for each column). data have; input col1 col2 col3 col4 col5 col6 col7; cards; 41 42 43 44 45 46 47 51 52 53 54 55 57 . 61 62 63 64 . . . 71 . . . . . . 81 82 83 84 . . . ; I tried the following code, as well as some other modifications, but can't quite get the correct answer. The following code does not include a space and includes "." for columns with missing values ("61626364..."). data want; set have; all=catt(' ', of col:); drop col:; run; Any advice?

luch25 · ‎08-27-2020

thank you @Reeza!

luch25 · ‎08-27-2020

I've written code using proc transpose (below), but it needs a modification that I can't quite figure out. data a; input id location $; cards; 1 A 1 B 1 C 2 A 3 A 3 B ; proc sql; create table want1 as select distinct id, location from a; run; proc transpose data = want1 out=want1a(drop =_name_) prefix = location_; by id; id location; var location; run; The resulting output gives me the character value for the location (A, B, C) for each location_X variable. In other words, I get: id location_A location_B location_C 1 A B C 2 A 3 A B The information is correct, but instead of a character value, I wanted a binary indicator 1/0. So essentially, I'm looking for: id location_A location_B location_C 1 1 1 0 2 1 0 0 3 1 1 0 Is there a way to modify my proc transpose code to get this? Or another way to do this?

luch25 · ‎08-27-2020

Got it, thank you for the explanation and help with the code! Have a great day!

luch25 · ‎08-27-2020

@FreelanceReinh this is exactly what I needed, thank you! Question- what does the "array popcnt[50]" do? Is this basically creating a rule for theoretical number of 50 states? So if I only had 8 states, could this be changed to popcnt[8]? That seemed to be the case when I played around with your code, but want to make sure.

luch25 · ‎08-27-2020

Yes, because some companies are located in multiple states (example: company D is located in states 5, 6, 7, 8). I calculated the weighted average in Excel, but the steps are basically: 1. Calculate the total population across all states (3000) 2. Calculate the state population for each company (e.g., company D has 4 states, with a total population of 1600) 3. Multiply the company's unweighted average by the: (value from step 2 / value from step 1). Do this for each company, and then sum across all companies for the weighted average.

luch25 · ‎08-27-2020

I have two different data sets that I need to use to calculate a weighted average: 1. Data set with population count for each state data state; input state pop_count; datalines; 1 200 2 300 3 400 4 500 6 200 7 300 8 500 ; 2. Dataset unweighted average for each company and states that they are located in. Some companies are located in multiple states, which is indicated by the colX variables. data company; input name $ mean col1 col2 col3 col4; datalines; A 0.6 1 . . . B 0.8 2 3 . . C 0.7 4 . . . D 0.5 5 6 7 8 ; The issue that I'm having is how to account for companies located in multiple states. How can I get the state-level total population count data from the state data set into the company data set? The final weighted average should be 0.61 for the data listed above.

Re: Select last value in calendar year

Select last value in calendar year

Re: Calculating averages for time intervals with different start dates

Re: Calculating averages for time intervals with different start dates

Re: Calculating averages for time intervals with different start dates

Re: Calculating averages for time intervals with different start dates

Calculating averages for time intervals with different start dates

Re: Count of unique values output as new variable

Count of unique values output as new variable

Re: Averaging multiple measurements within time interval

Re: Select last value in calendar year

Re: Select last value in calendar year

Re: Select last value in calendar year

Re: Count of unique values output as new variable

Re: Count of unique values output as new variable

Count of unique values output as new variable

Re: Modification to proc transpose output

Creating time intervals for longitudinal data, and taking averages wit...

Re: Creating date-based arrays for longitudinal data

Re: Creating date-based arrays for longitudinal data

Re: Creating date-based arrays for longitudinal data

Creating date-based arrays for longitudinal data

Re: Proc sql count for multiple variables

Proc sql count for multiple variables

Re: Concatenate numeric column values

Concatenate numeric column values

Re: Modification to proc transpose output

Modification to proc transpose output

Re: Calculate weighted average data from two files

Re: Calculate weighted average data from two files

Re: Calculate weighted average data from two files

Calculate weighted average data from two files