I have a table as below
level1 level2
furniture chair
furniture chair
furniture table
food pizza
I want to get unique level 1 variable and highest occurring level 2 variable for that variable. Output:
level1 level2
furniture chair
food pizza
Summarize data to get counts, PROC FREQ.
Sort into order to get the counts descending by level1.
Use a data step and BY group processing to get the first observations for each level1.
proc freq data=have order = freq;
table level1*level2 / out=counts;
run;
proc sort data=counts;
by level1 descending Count;
run;
data want;
set counts;
by level1;
if first.level1;
run;
Related
Noob SAS user here.
I have a hospital data set with patientID and a variable that counts the days between admission and discharge.
Those patients who had more than one hospital admission show up with the same patientID and with a record of how many days they were in hospital each time.
I want to sum the total days in hospital per patient, and then only have one patientID record with the sum of all hospital days across all stays. Does anyone know how I would go about this?
You want to select distinct the sum of days_in_hospital and group by patientID This will get what you want:
proc sql;
create table want as
select distinct
patientID,
sum(days_in_hospital) as sum_of_days
from have
group by patientID;
quit;
Alternatively you can use proc summary.
proc summary data= hospital_data nway;
class patientID;
var days;
output out=summarized_data (drop = _type_ _freq_) sum=;
run;
This creates a new dataset called summarized_data which has the summed days for each patientID. (The nway option removes the overall summary row, and the drop statement removes extra default summary columns you don't need.)
I want to compute multiple sums on the same column based on some criteria. Here is a small example using the sashelp.cars dataset.
The code below somewhat achieves what I want to do in three (3) different ways, but there is always a small problem.
proc report data=sashelp.cars out=test2;
column make type,invoice type,msrp;
define make / group;
define type / across;
define invoice / analysis sum;
define msrp / analysis sum;
title "Report";
run;
proc print data=test2;
title "Out table for the report";
run;
proc summary data=test nway missing;
class make type;
var invoice msrp;
output out=sumTest(drop= _Freq_ _TYPE_) sum=;
run;
proc transpose data=sumTest out=test3;
by make;
var invoice msrp;
id type;
run;
proc print data=test3;
title "Table using proc summary followed by proc transpose";
run;
proc sql undo_policy=none;
create table test4 as select
make,
sum(case when type='Sedan' then invoice else 0 end) as SedanInvoice,
sum(case when type='Wagon' then invoice else 0 end) as WagonInvoice,
sum(case when type='SUV' then invoice else 0 end) as SUVInvoice,
sum(case when type='Sedan' then msrp else 0 end) as Sedanmsrp,
sum(case when type='Wagon' then msrp else 0 end) as Wagonmsrp,
sum(case when type='SUV' then msrp else 0 end) as SUVmsrp
from sashelp.cars
group by make;
quit;
run;
proc print data=test4;
title "Table using SQL queries and CASE/WHEN to compute new columns";
run;
Here is the result I get when I run the presented code.
The first two tables represent the result and the out table of the report procedure. The problem I have with this approach is the column names produced by proc report. I would love to be able to define them myself, but I don't see how I can do this. It is important for further referencing.
The third table represent the result of the proc summary/proc transpose portion of the code. The problem I have with this approach is that Invoice and MSRP appears as rows in the table, instead of columns. For that reason, I think the proc report is better.
The last table represents the use of an SQL query. The result is exactly what I want, but the code is heavy. I have to do a lot of similar computation on my dataset and I believe this approach is cumbersome.
Could you help improve one of these methods ?
You can just use two PROC TRANSPOSE steps;
proc summary data=sashelp.cars nway missing;
where make=:'V';
class make type;
var invoice msrp;
output out=step1(drop= _Freq_ _TYPE_) sum=;
run;
proc transpose data=step1 out=step2;
by make type ;
var invoice msrp;
run;
proc transpose data=step2 out=step3(drop=_name_);
by make;
id type _name_ ;
var col1 ;
run;
proc print data=step3;
title "Table using proc summary followed by 2 proc transpose steps";
run;
Results:
Sedan Sedan Wagon Wagon
Obs Make SUVInvoice SUVMSRP Invoice MSRP Invoice MSRP
1 Volkswagen $32,243 $35,515 $335,813 $364,020 $77,184 $84,195
2 Volvo $38,851 $41,250 $313,990 $333,240 $57,753 $61,280
Use Proc TABULATE. Very succinct expressions for specifying row and column dimensions defined by desired hierarchy of class variables.
The intersection of these dimensions is a cell and represents a combination of values that select the values for which a statistical measure is displayed in the cell.
In your case the SUM is sum of dollars, which might not make sense when the cell has more then one contributing value.
For example: Does it make sense to show the invoice sum for 11 Volkswagen Sedan's is $335,813 ?
Also note the 'inverted' hierarchy used to show the number of contributing values.
Example:
proc tabulate data=sashelp.cars;
class make type;
var invoice msrp;
table
make=''
,
type * invoice * sum=''*f=dollar9.
type * msrp * sum=''*f=dollar9. /* this is an adjacent dimension */
(invoice msrp) * type * n='' /* specify another adjacent dimension, with inverted hierarchy */
/
box = 'Make'
;
where make =: 'V';
run;
Output
I have the following dataset :
data test;
input business_ID $;
datalines;
'busi1'
'busi1'
'busi1'
'busi2'
'busi3'
'busi3'
;
run;
proc freq data = test ;
table business_ID;
run;
I would like the average nummber of lines per business, that is count the total number of observations and divide it by the number of distinct businesses.
In my example : 6 observations, 3 businesses -> 6/2=3 lines per business.
I was thinking about using a proc freq or a proc mean step but so far I got only the number of lines (~freq) per business and do not know how to get to my goal.
Any idea?
You could use PROC FREQ to get the counts and then run PROC MEANS on the output.
proc freq data=test ;
tables business_id / noprint out=counts ;
run;
proc means data=counts;
var count;
run;
Or you could count them directly with PROC SQL code.
proc sql ;
select count(*)/count(distinct business_id) as mean_count
from test
;
quit;
Here is my data :
data example;
input id sports_name;
datalines;
1 baseball
1 basketball
1 cricket
1 soccer
2 golf
2 fencing
This is just a sample. The variable sports_name is categorical with 56 types.
I am trying to transpose the data to wide form where each row would have a user_id and the names of sports as the variables with values being 1/0 indicating Presence or absence.
So far, I used proc freq procedure to get the cross tabulated frequency table and put that in a different data set and then transposed that data. Now i have missing values in some cases and count of the sports in rest of the cases.
Is there any better way to do this?
Thanks!!
You need a way to create something from nothing. You could have also used the SPARSE option in PROC FREQ. SAS names cannot have length greater than 32.
data example;
input id sports_name :$16.;
retain y 1;
datalines;
1 baseball
1 basketball
1 cricket
1 soccer
2 golf
2 fencing
;;;;
run;
proc print;
run;
proc summary data=example nway completetypes;
class id sports_name;
output out=freq(drop=_type_);
run;
proc print;
run;
proc transpose data=freq out=wide(drop=_name_);
by id;
var _freq_;
id sports_name;
run;
proc print;
run;
Same theory here, generate a list of all possible combinations using SQL instead of Proc Summary and then transposing the results.
data example;
informat sports_name $20.;
input id sports_name $;
datalines;
1 baseball
1 basketball
1 cricket
1 soccer
2 golf
2 fencing
;
run;
proc sql;
create table complete as
select a.id, a_x.sports_name, case when not missing(e.sports_name) then 1 else 0 end as Present
from (select distinct ID from example) a
cross join (select distinct sports_name from example) a_x
full join example as e
on e.id=a.id
and e.sports_name=a_x.sports_name;
quit;
proc transpose data=complete out=want;
by id;
id sports_name;
var Present;
run;
Hell everyone --
I have some sales data which looks like this:
data have;
input order_id item $;
cards;
1 A
1 B
2 A
2 C
3 B
4 A
4 B
;
run;
What I'm trying to find out is what are the most popular combinations of items ordered. For example in the above case, there were 2 orders that contained items A&B, 1 order of A&C, and 1 order of B. What would be the best way to output the different combinations along with the numbers of orders placed?
It seems there is no permutation issue, you could try this:
proc sort data=have;
by order_id item;
run;
data temp;
set have;
by order_id;
retain comb;
length comb $4;
comb=cats(comb,item);
if last.order_id then do;
output;
call missing(comb);
end;
run;
proc freq data=temp;
table comb/norow nopercent nocol nocum;
run;
There are many possible approaches to this problem, and I would not presume to say which is the best. Here's a fairly simple method you could use:
Transpose your data so that you only have 1 row for each order, with an indicator variable for each product.
Feed the transposed dataset into proc corr to produce a correlation matrix for the indicator variables, and look for the strongest correlations.