topic Re: Creating a 5 level categorical variable. in Statistical Procedures

Creating a 5 level categorical variable.

saza — Mon, 31 Jan 2022 21:43:43 GMT

I was asked to make a 5 categorical variable and was able to find another forum that answered questions about making one, however, am having trouble applying it to my code. The first question is the following:

 Use the UNIVARIATE procedure to find the quintile (5-level) cut-points for the variable WEIGHT.
Using these cut-points, create a 5-level categorical variable called WTQUINT in a dataset called
MOD4_1 created using the TEMP6 dataset we left off with earlier. Run a FREQ procedure on this
variable.;

The following is my code that I thought was ok:

PROC UNIVARIATE DATA=temp6;
VAR WEIGHT;
RUN; 

DATA MOD4_1;
SET Temp6;
IF 0 < WEIGHT <= 135 THEN WTQUINT=0;
IF WEIGHT > 135 THEN WTQUINT=1;
RUN;

PROC FREQ;
TABLES WTQUINT;
RUN;

PROC UNIVARIATE DATA=MOD4_1;
VAR WEIGHT;
OUTPUT OUT=CUTPTS PCTLPTS= .20 .40 .60 .80 PCTLPRE = P_;
RUN; 

PROC PRINT DATA= CUTPTS;
RUN;

The second question is the following:

*2. Use the RANK procedure to create a 5-level categorical variable called WTQUINT2 in a dataset called
MOD4_2 from the dataset MOD4_1.

How do I create a ranking system if my obs values are 94, 94, 95, 95? I'm guessing I messed up when making my 5 categorical variable. I took 1/5, 2/5, 3/5, 4/5 out of 100% to get .20, .40. .60, .80 in my code.

Re: Creating a 5 level categorical variable.

StatDave — Mon, 31 Jan 2022 21:50:55 GMT

Use the GROUPS=x option in PROC RANK to create a variable with x levels based on percentiles.

Re: Creating a 5 level categorical variable.

saza — Mon, 31 Jan 2022 21:58:25 GMT

PROC RANK DATA=mod4_2 OUT=TWO GROUPS=X;
VAR WEIGHT;
RANKS WTQUINT2;
RUN;

this?

Re: Creating a 5 level categorical variable.

PaigeMiller — Mon, 31 Jan 2022 22:09:40 GMT

Well no, you don't use X, you use the actual number of groups (or levels) you want.

Re: Creating a 5 level categorical variable.

StatDave — Mon, 31 Jan 2022 22:11:01 GMT

Specify GROUPS=5

Re: Creating a 5 level categorical variable.

saza — Mon, 31 Jan 2022 22:18:29 GMT

Yes I can see that. I guess what I should clarify is that i'm getting blank results when I use the code:
PROC RANK DATA=mod4_2 OUT=TWO GROUPS=5;
VAR WEIGHT;
RANKS WTQUINT2;
RUN;

Because my obs values from the first part of the question is
P_0_2= 94
P_0_4= 94
P_0_6=95
P_0_8=95

So my ranking isnt working.

Because I had to do the following before ranking:

DATA MOD4_2;
SET MOD4_1;
IF WEIGHT > 0 AND WEIGHT <= 94 THEN WTQUINT2=0;
IF WEIGHT > 94 AND WEIGHT <= 95 THEN WTQUINT2=1;
RUN;

Re: Creating a 5 level categorical variable.

StatDave — Mon, 31 Jan 2022 22:19:44 GMT

It doesn't produce printed output. Look at the contents of the created data set - TWO

Re: Creating a 5 level categorical variable.

saza — Mon, 31 Jan 2022 22:24:36 GMT

So,
PROC RANK DATA=mod4_2 OUT=FOUR GROUPS=5;
VAR WEIGHT;
RANKS WTQUINT2;
RUN;
PROC FREQ DATA=mod4_2;
TABLES WTQUINT2;
RUN;

I still get a blank table

Re: Creating a 5 level categorical variable.

StatDave — Mon, 31 Jan 2022 22:26:50 GMT

Look at the OUT= data set NOT the input data set (mod4_2)
proc freq data=four;

Re: Creating a 5 level categorical variable.

saza — Mon, 31 Jan 2022 22:30:43 GMT

Sorry, as you can tell I'm very new to SAS.

So Proc Freq data=four; where do I place that in my code?

Re: Creating a 5 level categorical variable.

StatDave — Mon, 31 Jan 2022 22:33:36 GMT

anywhere after the input data set (mod4_2) is created.

Re: Creating a 5 level categorical variable.

saza — Mon, 31 Jan 2022 22:41:03 GMT

Wish I could post an image of my blank table. I believe it is how I did my > and < steps prior. But your code is working.

Re: Creating a 5 level categorical variable.

Tom — Mon, 31 Jan 2022 23:24:50 GMT

I think they are asking you to use PROC RANK on your original dataset.

So no need for you to manually try to create the 5 categories, let PROC RANK do that for you.

Re: Creating a 5 level categorical variable.

saza — Tue, 01 Feb 2022 00:14:28 GMT

No I figured it out. It was because I put a . infront of .20 instead of just 20 bc sas is already reading it as .20