turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- SAS Programming
- /
- General Programming
- /
- SAS_Count Frequency

Topic Options

- RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

07-01-2017 01:39 PM - edited 07-01-2017 01:41 PM

Hi all,

I want to ask a complicated (for me) question about SAS programming. I think I can explain better by using simple example. So, I have the following dataset:

Group Category

A 1

A 1

A 2

A 1

A 2

A 3

B 1

B 2

B 2

B 1

B 3

B 2

I want to count the each category for each group. I can do it by using PROC FREQ. But it is not better way for my dataset. It will be time consuming for me as my dataset is too large and I have a huge number of groups. So, if I use PROC FREQ, firstly I need to create new datasets for each group and then use PROC FREQ for each group. In sum, I need to create the following dataset:

CATEGORIES

Group 1 (first category) 2 3

A 3 (the number of first category in A group is 3) 2 1

B 2 (the number of first category in B group is 2) 3 1

I do not write explainations for second and third categories as it is the same with the first category.I think I can explain it. Thanks for your helps.

Accepted Solutions

Solution

07-02-2017
04:36 AM

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Khaladdin

07-01-2017 02:50 PM

If it takes too much time or you are short with memory, add a sort:

```
proc sort data=have; by group; run;
proc freq data=have;
by group;
table group*category / nopercent
out=want;
run;
```

All Replies

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Khaladdin

07-01-2017 02:26 PM

You wrote that you have " a huge number of groups". How huge is it ?

Try to run:

```
proc freq data=have;
table group*category / nopercent
out=want;
run;
```

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Shmuel

07-01-2017 02:43 PM

Thanks for reply. It is high frequency data. So, I have more than one million groups.

Solution

07-02-2017
04:36 AM

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Khaladdin

07-01-2017 02:50 PM

If it takes too much time or you are short with memory, add a sort:

```
proc sort data=have; by group; run;
proc freq data=have;
by group;
table group*category / nopercent
out=want;
run;
```

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Khaladdin

07-01-2017 02:53 PM

Many thanks. I will try it tomorrow. I will let you know about the result. Many thanks agaib.

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Khaladdin

07-01-2017 04:44 PM

It appears to me that you want, for each group, to rank categories for each group.

Then produce one obs for each group with variables CAT1, CAT2, .... CATj. CAT1 will contain the most frequent category (what I think you mean by "first category"), CAT2 the second most frequent, through CATj (where J is the number of distinct categories in the group having the highest cardinatliy.

If your data are aleady sorted by group, it's two steps:

- Use proc freq to generate a data set T1 of frequencies sorted by GROUP and then descending frequency ("order=freq") within group.

`proc freq data=have noprint order=freq; by group; table category / out=t1; run;`

- Then just use proc transpose to collapse the descending frequencies for each group into a single row with vars CAT1, CAT2, ...

`proc transpose data=t1 out=want prefix=CAT; by group; var category; run;`

Now if your data are NOT sorted, it's three steps - but don't worry - you don't have to sort the original data, just the frequencies:

- (and 2) Get frequencies of group*category and sort by descending frequency within each group:
`proc freq data=have noprint ; table group * category / out=t1; run; proc sort data=t1; by group descending count; run;`

- (see above)
- Now transpose a column to a row for each group.

`proc transpose data=t1 prefix=CAT; by group; var category; run;`

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to mkeintz

07-02-2017 04:37 AM

Frequency works. But, transpose does not work.

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Khaladdin

07-05-2017 09:54 PM

Show the log of the non-working code please.

- Mark as New
- Bookmark
- Subscribe
- RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Khaladdin

07-02-2017 01:47 PM

Hi, another idea (not 1 million groups, but 12 million observations ... SUMMARY + TRANSPOSE juts a few seconds). I don't see in your exmaple output the need to rank counts within groups.

**data x;****input (group category) (:$1.) @@;****datalines;****A 1 A 1 A 2 A 1 A 2 A 3****B 1 B 2 B 2 B 1 B 3 B 2****; **

**data bigx;****set x;****do _n_=1 to 1e6;****output;****end;****run;**

**proc summary data=bigx nway;****class group category;****output out=y (drop=_type_); ****run;**

**proc transpose data=y out=z (drop=_name_) prefix=cat;****by group;****var _freq_;****id category;****run;**

**DATA SET: z**

**group cat1 cat2 cat3**

**A 3000000 2000000 1000000**** B 2000000 3000000 1000000**