BookmarkSubscribeRSS Feed
mandan414
Fluorite | Level 6

Hello,

 

I'm estimating a series of cross tabulation using a large dataset. There are a few threshold categories - based on the sample counts- that needs to be considered for data release. For example, like this:

 

- category A (good quality):  we use the total population and the sample counts is >=500.  we use the subpopulation of age >=20 and the sample counts is >=700.

- category E (marginal quality):  we use the total population and the sample counts is between 250 and 500 .  we use the subpopulation of age >=20 and the sample counts is between 350 and 700.

- category F (poor quality):  we use the total population and the sample counts is <250. we use the subpopulation of age >=20 and the sample counts is <350.

There are many cells from cross -tabulations and it is very time consuming to check every cell for the thresholds. is there any ways to categorize the output (proportions) according to the above threshold?

 

Thank you.

5 REPLIES 5
PaigeMiller
Diamond | Level 26

I find this impossible to understand without a small example. Show us realistic example data of your data, and what the desired output is.

--
Paige Miller
ballardw
Super User

Example data - before the cross tabulation

How you make the cross tab

What is the content of the "many cells"?

 

Your "rules" don't make much sense. Basing a category on a fixed number of counts is, in my mind, unreliable because your basic sample size might change. If the total sample size is fixed, then tell us what that number must be.

mandan414
Fluorite | Level 6

Thank you so much to "PaigeMiller" and Bllardw" for their responses.

My total sample size is fixed. I'm estimating the proportions of people for 12 categorical indicators disaggregated by sex, age, education .

 

Here is the example of data:

variable name: indicate 1

Answer categories            code

great                                    1

moderate                             2

small                                    3

not at all                               4

 

variable name: indicate 2

Answer categories            code

great                                    1

moderate                             2

small                                    3

not at all                               4

.

.

.

variable name: indicate 12

Answer categories            code

great                                    1

moderate                             2

small                                    3

not at all                               4

 

an example of my syntax is (I'm using SUDAAN in SAS)

proc crosstab data=mydata design=BBR;

weight    myweight;

repwgt    mybootstrapweight;

tables     gender * indicate1;

class       gender indicate1;

setenv     colwidth = 1;

output      nsum   serow  rowper lowrow uprow /filename= out_   filetype=SAS;

 

data out;

set out_;

row= round (rowper, .1);

lowR= round (lowrow, .1);

upR= round (uprow, .1);

cv= round ((serow/rowper *100), .1);

keep gender indicate 1  nsum row lowR upR cv;

run;

I need to marked the output of "nsum" based on the thresholds, for example nsum>=220 be marked as  A (good quality), 110>=nsum>=220 be marked as E (marginal quality) and nsum<110 be marked as F for poor quality.

 

Hope I could explain it well now.

Thank you so much.

 

 

 

PaigeMiller
Diamond | Level 26

An example of your output was specifically requested, but you didn't provide that. (For those of us who don't understand SUDAAN, your explanation doesn't help.)

 

It also seems that this is not a small example of the data that we asked for, it is just a description of your variables and what the values mean, which isn't sufficient, we need to see a small portion of the actual SAS data set you will be analyzing, in the format requested.

 

 

--
Paige Miller
ballardw
Super User

IF this is actually correct:

data out;
set out_;
row= round (rowper, .1);
lowR= round (lowrow, .1);
upR= round (uprow, .1);
cv= round ((serow/rowper *100), .1);
keep gender indicate 1  nsum row lowR upR cv;
run;

I need to marked the output of "nsum" based on the thresholds,
for example nsum>=220 be marked as A (good quality),
110>=nsum>=220 be marked as E (marginal quality) and
nsum<110 be marked as F for poor quality.

Then perhaps a custom format is a possible solution. HOWEVER you need to pick ONE category for the value 220. What you show has 220 in both A and E.

Here is one way with the 220 value in category E:

Proc format;
value MyCategory
0   -<110 = 'F'
110 - 220 = 'E'
220<- high= 'A'
;
run;

proc print data=out noobs;
var nsum;
format nsum MyCategory.;
run;

Formats are a very powerful tool in SAS. Consider if someone looks at your output and says what to things look like if set 150 for the boundary between F and E? Or wants a fourth result added. You create a different format for use at report or analysis time instead of creating new variables. The custom formats will work for all the report and most of the analysis and graphing procedures.

 

sas-innovate-2024.png

Available on demand!

Missed SAS Innovate Las Vegas? Watch all the action for free! View the keynotes, general sessions and 22 breakouts on demand.

 

Register now!

How to Concatenate Values

Learn how use the CAT functions in SAS to join values from multiple variables into a single value.

Find more tutorials on the SAS Users YouTube channel.

Click image to register for webinarClick image to register for webinar

Classroom Training Available!

Select SAS Training centers are offering in-person courses. View upcoming courses for:

View all other training opportunities.

Discussion stats
  • 5 replies
  • 453 views
  • 0 likes
  • 3 in conversation