I have a dataset like below
Customer_ID Vistited_Date
1234 7-Feb-20
4567 7-Feb-20
9870 7-Feb-20
1234 14-Feb-20
7654 14-Feb-20
3421 14-Feb-20
I am trying find the cumulative unique count of customers by date, assuming my output will be like below
Cust_count Vistited_Date
3 7-Feb-20
2 14-Feb-20
7-Feb-2020 has 3 unique customers, whereas 14-Feb-2020 has only 2 hence customer 1234 has visited already.
Anyone knows how I could develop a data set in these conditions?
Sorry if my question is not clear enough, and I am available to give more details if necessary.
Thanks!
NOTE: #draycut's answer has the same logic but is faster, and I will explain why.
#draycut's code uses one hash method, add(), using the return code as test for conditional increment. My code uses check() to test for conditional increment and then add (which will never fail) to track. The one method approach can be perceived as being anywhere from 15% to 40% faster in performance (depending on number of groups, size of groups and id reuse rate)
You will need to track the IDs that have occurred in all prior groups, and exclude the tracked IDs from the current group count.
Tracking can be done with a hash, and conditional counting can be performed in a DOW loop over each group. A DOW loop places the SET statement inside an explicit DO.
Example:
data have;
input ID Date: date9.; format date date11.;
datalines;
1234 7-Feb-20
4567 7-Feb-20
9870 7-Feb-20
1234 14-Feb-20
7654 14-Feb-20
3421 14-Feb-20
;
data counts(keep=date count);
if _n_ = 1 then do;
declare hash tracker();
tracker.defineKey('id');
tracker.defineDone();
end;
do until (last.date);
set have;
by date;
if tracker.check() ne 0 then do;
count = sum(count, 1);
tracker.add();
end;
end;
run;
Raw performance benchmark - no disk io, cpu required to fill array before doing hashing, so those performance components are combined.
The root performance is how fast can new items be added to the hash.
Simulate 3,000,000 'records', 1,000 groups of 3,000 dates, 10% id reuse (so the distinct ids will be ~2.7M).
%macro array_fill (top=3000000, n_group = 1000, overlap_factor=0.10);
%local group_size n_overlap index P Q;
%let group_size = %eval (&top / &n_group);
%if (&group_size < 1) %then %let group_size = 1;
%let n_overlap = %sysevalf (&group_size * &overlap_factor, floor);
%if &n_overlap < 0 %then %let n_overlap = 0;
%let top = %sysevalf (&group_size * &n_group);
P = 1;
Q = &group_size;
array ids(&top) _temporary_;
_n_ = 0;
do i = 1 to &n_group;
do j = P to Q;
_n_+1;
ids(_n_) = j;
end;
P = Q - &n_overlap;
Q = P + &group_size - 1;
end;
%mend;
options nomprint;
data _null_ (label='check then add');
length id 8;
declare hash h();
h.defineKey('id');
h.defineDone();
%array_fill;
do index = 1 to dim(ids);
id = ids(index);
if h.check() ne 0 then do;
count = sum(count,1);
h.add();
end;
end;
_n_ = h.num_items;
put 'num_items=' _n_ comma12.;
put index= comma12.;
stop;
run;
data _null_ (label='just add');
length id 8;
declare hash h();
h.defineKey('id');
h.defineDone();
%array_fill;
do index = 1 to dim(ids);
id = ids(index);
if h.add() = 0 then
count = sum(count,1);
end;
_n_ = h.num_items;
put 'num_items=' _n_ comma12.;
put index= comma12.;
stop;
run;
data have;
input Customer_ID Vistited_Date :anydtdte12.;
format Vistited_Date date9.;
datalines;
1234 7-Feb-2020
4567 7-Feb-2020
9870 7-Feb-2020
1234 14-Feb-2020
7654 14-Feb-2020
3421 14-Feb-2020
;
data want (drop=Customer_ID);
if _N_=1 then do;
declare hash h ();
h.definekey ('Customer_ID');
h.definedone ();
end;
do until (last.Vistited_Date);
set have;
by Vistited_Date;
if h.add() = 0 then Count = sum(Count, 1);
end;
run;
If your data is not sorted and you like the SQL maybe this solution is same good for you and it is very simple:
/* your example 3 rows */
data have;
input ID Date: date9.; format date date11.;
datalines;
1234 7-Feb-20
4567 7-Feb-20
9870 7-Feb-20
1234 14-Feb-20
7654 14-Feb-20
3421 14-Feb-20
1234 15-Feb-20
7654 15-Feb-20
1111 15-Feb-20
;
run;
/* simple set theory. Final dataset contains your final data like results
below*/
proc sql;
create table temp(where =(mindate=date)) as select
ID, date,min(date) as mindate from have
group by id;
create table final as select count(*) as customer_count,date from temp
group by date;
quit;
/* results:
customer_count Date
3 07.febr.20
2 14.febr.20
1 15.febr.20
*/
Another method cause I dont know hash so well. >_<
data have;
input ID Date: date9.; format date date11.;
datalines;
1234 7-Feb-20
4567 7-Feb-20
9870 7-Feb-20
1234 14-Feb-20
7654 14-Feb-20
3421 14-Feb-20
;
data want;
length Used $200.;
retain Used;
set have;
by Date;
if first.Date then count = .;
if not find(Used,cats(ID)) then do;
count + 1;
Used = catx(',',Used,ID);
end;
if last.Date;
put Date= count=;
run;
If you are not overly concerned with processing speed and want something simple:
proc sort data=have;
by id date;
** Get date of each customer's first unique visit **;
proc sort data=have out=first_visit nodupkey;
by id;
proc freq data=first_visit noprint;
tables date /out=want (keep=date count);
run;
Related
Since I am new to SAS I need some help to understand how to combine the overlap date ranges into one row.I want to combine the overlap date ranges when they have matching Id. If the dates don’t overlap then I want to keep them as it is. IF they over lap by Matching Id and drug code Then it should combine into one line. Please look at the same ple data set which I have below and the expected results:
Current Data set:
ID Drug Code BEG_Date End_Date
1 100 1/1/2018 1/1/2019
1 100 1/1/2018 3/1/2018
1 100 2/1/2018 04/30/2018
1 90 4/1/2018 04/30/2018
1 100 5/1/2018 6/1/2018
1 98 6/1/2018 8/31/2018
1 100 9/1/2018 5/4/2019
Expected results:
ID Drug Code BEG_Date End_Date
1 100 1/1/2018 3/31/2018
1 90 4/1/2018 04/30/2018
1 100 5/1/2018 6/1/2018
1 98 6/2/2018 8/31/2018
1 100 9/1/2018 5/4/2019
I wrote some SAS code but I am combining the dates even when there is no overlap. I want to write some code which should work in SAS.
PROC SORT DATA=Want OUT=ONE;
BY PERSON_ID BEG_DATE DRUG_CODE END_DATE;
RUN;
data TWO (DROP=PERSON_ID2 DRUG_CODE2 BEG_DATE END_DATE
RENAME=(BEG2=BEG_DOS
END2=END_DOS));
SET ONE;
RETAIN BEG2 END2;
PERSON_ID2=LAG1(PERSON_ID);
DRUG_CODE2=LAG1(DRUG_CODE);
IF PERSON_ID2=PERSON_ID AND DRUG_CODE2=DRUG_CODE AND BEG_DATE LE(END2+1) THEN
DO;
BEG2=MIN(BEG_DATE,BEG2);
END2=MAX(END_DATE,END2);
END;
ELSE
DO;
SEG+1;
BEG2=BEG_DATE;
END2=END_DATE;
END;
FORMAT BEG2 END2 MMDDYY10.;
RUN;
DATA THREE(DROP=BEG_DOS END_DOS SEG);
RETAIN BEG_DATE END_DATE;
SET TWO;
BY PERSON_ID SEG;
FORMAT BEG_DATE END_DATE MMDDYY10.;
IF FIRST.SEG THEN
DO;
BEG_DATE=BEG_DOS;
END;
IF LAST.SEG THEN
DO;
END_DATE = END_DOS;
OUTPUT;
END;
RUN;
This is how I would do it. Create an obs for each ID DRUG and DATE. Flag the gaps and summarize by RUN.
data have;
input ID Drug_Code (BEG End)(:mmddyy.);
format BEG End mmddyyd10.;
cards;
1 100 1/1/2018 3/1/2018
1 100 2/1/2018 04/30/2018
1 90 4/1/2018 04/30/2018
1 90 6/1/2018 8/15/2018
1 100 5/1/2018 6/1/2018
1 98 6/1/2018 8/31/2018
1 100 9/1/2018 5/4/2019
;;;;
run;
proc print;
run;
/*1 100 1/1/2018 1/1/2019*/
data exv/ view=exv;
set have;
do date = beg to end;
output;
end;
drop beg end;
format date mmddyyd10.;
run;
proc sort data=exv out=ex nodupkey;
by id drug_code date;
run;
data breaksV / view=BreaksV;
set ex;
by id drug_code;
dif = dif(date);
if first.drug_code then do; dif=1; run=1; end;
if dif ne 1 then run+1;
run;
proc summary data=breaksV nway missing;
class id drug_code run;
var date;
output out=want(drop=_type_) min=Begin max=End;
run;
Proc print;
run;
Computing the extent range composed of overlapping segment ranges requires a good understanding of the range conditions (cases).
Consider the scenarios when sorted by start date (within any larger grouping set, G, such as id and drug)
Let [ and ] be endpoints of a range
# be date values (integers) within
Extent be the combined range that grows
Segment be the range in the current row
Case 1 - Growth. Within G Segment start before Extent end
Segment will either not contribute to Extent or extend it.
[####] Extent
+ [#] Segment range DOES NOT contribute
--------
[####] Extent (do not output a row, still growing)
or
[####] Extent
+ [#####] Segment range DOES contribute
--------
[#######] Extent (do not output a row, still growing)
Case 2 - Terminus. 3 possibilities:
Within G Segment start after Extent end,
Next G reached (different id/drug combination),
End of data reached.
#2 and #3 can be tested by checking the appropriate last. flag.
[####] Extent
+ ..[#] Segment beyond Extent (gap is 2)
--------
[####] output Extent
[#] reset Extent to Segment
You can adjust your rules for Segment being adjacent (gap=0) or close enough (gap < threshold) to mean an Extent is either expanded, or, output and reset to Segment.
Note: The situation is a little more (not shown) complicated for the real world cases of:
missing start means the Segment has an unknown start date (presume it to be epoch (0=01JAN1960, or some date that pre-dates all dates in the data or study)
missing end means the Segment is active today (end date is date when processing data)
Sample code:
data have;
call streaminit(42);
do id = 1 to 10;
do _n_ = 1 to 50;
drug = ceil(rand('UNIFORM', 10));
beg_date = intnx ('MONTH', '01JAN2008'D, rand('UNIFORM',20));
end_date = intnx ('DAY', beg_date, rand('UNIFORM',75));
OUTPUT;
end;
end;
format beg_date end_date yymmdd10.;
run;
proc sort data=have out=segments;
by id drug beg_date end_date;
run;
data want;
set segments;
by id drug beg_date end_date; * will error if incoming data is NOT sorted;
retain ext_beg ext_end;
retain gap_allowed 0; * set to 1 for contiguously adjacent segment ;
if first.drug then do;
ext_beg = beg_date;
ext_end = end_date;
segment_count = 0;
end;
if beg_date <= ext_end + gap_allowed then do;
ext_end = max (ext_end, end_date);
segment_count + 1;
end;
else do;
extent_id + 1;
OUTPUT;
ext_beg = beg_date;
ext_end = end_date;
segment_count = 1;
end;
if last.drug then do;
extent_id + 1;
OUTPUT;
* reset occurs implicitly;
* it will happen at first. logic when control returns to top of step;
end;
format ext_: yymmdd10.;
keep id drug ext_beg ext_end segment_count extent_id;
run;
I have a dataset of patient data with each diagnosis on a different line.
This is an example of what it looks like:
patientID diabetes cancer age gender
1 1 0 65 M
1 0 1 65 M
2 1 1 23 M
2 0 0 23 M
3 0 0 50 F
3 0 0 50 F
I need to isolate the patients who have a diagnosis of both diabetes and cancer; their unique patient identifier is patientID. Sometimes they are both on the same line, sometimes they aren't. I am not sure how to do this because the information is on multiple lines.
How would I go about doing this?
This is what I have so far:
PROC SQL;
create table want as
select patientID
, max(diabetes) as diabetes
, max(cancer) as cancer
, min(DOB) as DOB
from diab_dx
group by patientID;
quit;
data final; set want;
if diabetes GE 1 AND cancer GE 1 THEN both = 1;
else both =0;
run;
proc freq data=final;
tables both;
run;
Is this correct?
If you want to learn about data steps lookup how this works.
data pat;
input patientID diabetes cancer age gender:$1.;
cards;
1 1 0 65 M
1 0 1 65 M
2 1 1 23 M
2 0 0 23 M
3 0 0 50 F
3 0 0 50 F
;;;;
run;
data both;
do until(last.patientid);
set pat; by patientid;
_diabetes = max(diabetes,_diabetes);
_cancer = max(cancer,_cancer);
end;
both = _diabetes and _cancer;
run;
proc print;
run;
add a having statement at the end of sql query should do.
PROC SQL;
create table want as
select patientID
, max(diabetes) as diabetes
, max(cancer) as cancer
, min(age) as DOB
from PAT
group by patientID
having calculated diabetes ge 1 and calculated cancer ge 1;
quit;
You might find some coders, especially those coming from statistical backgrounds, are more likely to use Proc MEANS instead of SQL or DATA step to compute the diagnostic flag maximums.
proc means noprint data=have;
by patientID;
output out=want
max(diabetes) = diabetes
max(cancer) = cancer
min(age) = age
;
run;
or for the case of all the same aggregation function
proc means noprint data=have;
by patientID;
var diabetes cancer;
output out=want max= ;
run;
or
proc means noprint data=have;
by patientID;
var diabetes cancer age;
output out=want max= / autoname;
run;
I am trying to go through a list of dates and keep only the date range for dates that 5 or more occurrences and delete all others. The example I have is:
data test;
input dt dt2;
format dt dt2 date9.;
datalines;
20000 20001
20000 20002
20000 20003
21000 21001
21000 21002
21000 21003
21000 21004
21000 21005
;
run;
proc sort data = test;
by dt dt2;
run;
data check;
set test;
by dt dt2;
format dt dt2 date9.;
if last.dt = first.dt then
if abs(last.dt2 - first.dt) < 5 then delete;
run;
What I want returned is just one entry, if possible, but I would be happy with the entire appropriate range as well.
The one entry would be a table that has:
start_dt end_dt
21000 21005
The appropriate range is:
21000 21001
21000 21002
21000 21003
21000 21004
21000 21005
My code doesn't work as desired, and I am not sure what changes I need to make.
last.dt2 and first.dt are flags and can have value in (0,1), so condition abs(last.dt2 - first.dt) < 5 is always true.
Use counter variable to count records in group instead:
data check(drop= count);
length count 8;
count=0;
do until(last.dt);
set test;
by dt dt2;
format dt dt2 date9.;
count = count+1;
if last.dt and count>=5 then output;
end;
run;
I'm not sure why you are looking to use the last.dt2 and the first.dt within your delete function so I have turned it around to create your desired output:
data check2;
set test;
by dt ;
format dt dt2 date9.;
if last.dt then do;
if abs(dt2 - dt) >= 5 then output;
end;
run;
Of course, this will only work if your file is sorted on dt and dt2.
Hope this helps.
I have the following data where people in households are sorted by age (oldest to youngest):
data houses;
input HouseID PersonID Age;
datalines;
1 1 25
1 2 20
2 1 32
2 2 16
2 3 14
2 4 12
3 1 44
3 2 42
3 3 10
3 4 5
;
run;
I would like to calculate for each household the maximum age difference between consecutively aged people. So this example would give values of 5 (=25-20), 16 (=32-16) and 32 (=42-10) for households 1, 2 and 3 consecutively.
I could do this using lots of merges (i.e. extract person 1, merge with extract of person 2, and so on), but as there can be upto 20+ people in a household I'm looking for a much more direct method.
Here's a two pass solution. Same first step as the two solutions above, sort by age. In the second step keep track of max_diff per row, at the last record of HouseID output the results. This results in only two passes through the data.
proc sort data=houses; by houseid age;run;
data want;
set houses;
by houseID;
retain max_diff 0;
diff = dif1(age)*-1;
if first.HouseID then do;
diff = .; max_diff=.;
end;
if diff>max_diff then max_diff=diff;
if last.houseID then output;
keep houseID max_diff;
run;
proc sort data=houses; by houseid personid age;run;
data _t1;
set houses;
diff = dif1(age) * (-1);
if personid = 1 then diff = .;
run;
proc sql;
create table want as
select houseid, max(diff) as Max_Diff
from _t1
group by houseid;
proc sort data = house;
by houseid descending age;
run;
data house;
set house;
by houseid;
lag_age = lag1(age);
if first.houseid then age_diff = 0;
age_diff = lag_age - age;
run;
proc sql;
select houseid,max(age_diff) as max_age_diff
from house
group by houseid;
quit;
Working:
First sort the data set using houseid and descending Age.
Second data step will calculate difference between current age value (in PDV) and previous age value in PDV. Then, using sql procedure, we can get the max age difference for each houseid.
Just throwing one more into the mix. This one is a condensed version of Reeza's response.
/* No need to sort by PersonID as age is the only concern */
proc sort data = houses;
by HouseID Age;
run;
data want;
set houses;
by HouseID;
/* Keep the diff when a new row is loaded */
retain diff;
/* Only replace the diff if it is larger than previous */
diff = max(diff, abs(dif(Age)));
/* Reset diff for each new house */
if first.HouseID then diff = 0;
/* Only output the final diff for each house */
if last.HouseID;
keep HouseID diff;
run;
Here is an example using FIRST. and LAST. with one pass (after sort) through the data.
data houses;
input HouseID PersonID Age;
datalines;
1 1 25
1 2 20
2 1 32
2 2 16
2 3 14
2 4 12
3 1 44
3 2 42
3 3 10
3 4 5
;
run;
Proc sort data=HOUSES;
by houseid descending age ;
run;
Data WANT(keep=houseid max_diff);
format houseid max_diff;
retain max_diff age1 age2;
Set HOUSES;
by houseid descending age ;
if first.houseid and last.houseid then do;
max_diff=0;
output;
end;
else if first.houseid then do;
call missing(max_diff,age1,age2);
age1=age;
end;
else if not(first.houseid or last.houseid) then do;
age2=age;
temp=age1-age2;
if temp>max_diff then max_diff=temp;
age1=age;
end;
else if last.houseid then do;
age2=age;
temp=age1-age2;
if temp>max_diff then max_diff=temp;
output;
end;
Run;
I have two SAS data sets. The first is relatively small, and contains unique dates and a corresponding ID:
date dateID
1jan90 10
2jan90 15
3jan90 20
...
The second data set very large, and has two date variables:
dt1 dt2
1jan90 2jan90
3jan90 1jan90
...
I need to match both dt1 and dt2 to dateID, so the output would be:
id1 id2
10 15
20 10
Efficiency is very important here. I know how to use a hash object to do one match, so I could do one data step to do the match for dt1 and then another step for dt2, but I'd like to do both in one data step. How can this be done?
Here's how I would do the match for just dt1:
data tbl3;
if 0 then set tbl1 tbl2;
if _n_=1 then do;
declare hash dts(dataset:'work.tbl2');
dts.DefineKey('date');
dts.DefineData('dateid');
dts.DefineDone();
end;
set tbl1;
if dts.find(key:date)=0 then output;
run;
A format would probably work just as efficiently given the size of your hash table...
data fmt ;
retain fmtname 'DTID' type 'N' ;
set tbl1 ;
start = date ;
label = dateid ;
run ;
proc format cntlin=fmt ; run ;
data tbl3 ;
set tbl2 ;
id1 = put(dt1,DTID.) ;
id2 = put(dt2,DTID.) ;
run ;
Edited version based on below comments...
data fmt ;
retain fmtname 'DTID' type 'I' ;
set tbl1 end=eof ;
start = date ;
label = dateid ;
output ;
if eof then do ;
hlo = 'O' ;
label = . ;
output ;
end ;
run ;
proc format cntlin=fmt ; run ;
data tbl3 ;
set tbl2 ;
id1 = input(dt1,DTID.) ;
id2 = input(dt2,DTID.) ;
run ;
I don't have SAS in front of me right now to test it but the code would look like this:
data tbl3;
if 0 then set tbl1 tbl2;
if _n_=1 then do;
declare hash dts(dataset:'work.tbl2');
dts.DefineKey('date');
dts.DefineData('dateid');
dts.DefineDone();
end;
set tbl1;
date = dt1;
if dts.find()=0 then do;
id1 = dateId;
end;
date = dt2;
if dts.find()=0 then do;
id2 = dateId;
end;
if dt1 or dt2 then do output; * KEEP ONLY RECORDS THAT MATCHED AT LEAST ONE;
drop date dateId;
run;
I agree with the format solution, for one, but if you want to do the hash solution, here it goes. The basic thing here is that you define the key as the variable you're matching, not in the hash itself.
data tbl2;
informat date DATE7.;
input date dateID;
datalines;
01jan90 10
02jan90 15
03jan90 20
;;;;
run;
data tbl1;
informat dt1 dt2 DATE7.;
input dt1 dt2;
datalines;
01jan90 02jan90
03jan90 01jan90
;;;;
run;
data tbl3;
if 0 then set tbl1 tbl2;
if _n_=1 then do;
declare hash dts(dataset:'work.tbl2');
dts.DefineKey('date');
dts.DefineData('dateid');
dts.DefineDone();
end;
set tbl1;
rc1 = dts.find(key:dt1);
if rc1=0 then id1=dateID;
rc2 = dts.find(key:dt2);
if rc2=0 then id2=dateID;
if rc1=0 and rc2=0 then output;
run;