Hi,
I have been trying to recode the BMI values of my study participants into three different categories. However, things are not working our the way they're supposed to.
I've used the syntax below. Most of the recoding goes fine, however, there are two participants in my dataset who have a BMI of EXACTLY 25. According to my syntax, these people should be assigned a BMIcat-value of 0.5. When I run the syntax though, these people are assigned a BMIcat-value of 0.
How is this possible and how can I correct this??
DATA pp;
SET pp;
BMIcat=.;
IF 18.5
Dear Miriam,
This is SAS bug as mentioned above.
You had better round up the BMI by appropriate digit number.
And please confirm the variable "error" by run the following program, the variable has the smallest error.
data p;
input weight height;
cards;
64 160
;
run;
data pp;
set p;
bmi=weight/(height/100)**2;
error=bmi-25;
run;
data ppp;
set pp;
bmicat=1;
if bmi=. then bmicat=.;
else if 18.5
Dear Miriam,
Do you have more than one measurement for each individual in your data set?
Except for duplicate measurements as Mahdi mentioned, if you use the formula BMI=Weight in Kg/height2 in m to evaluate BMI, you may not get exact value of 25, say it might be 24.999999, but it was shown as 25.
I work in a personperiod dataset. Each subject has multiple rows, but each row has only one wieght and one height measurement value.
The persons who are recoded wrong have a BMI of exact 25 (weight 64 kg, height 1.6 m.).
Strangly the problem was fixed after I changed the BMI formula from
BMI = weight / (height/100)**2;
into:
BMI= weight/((height/100)*(height/100));
I don't understand why this appears to matter, but I'm happy the problem is solved for now...
Miriam,
This is a bug of SAS, you may write a email to SAS asking them to fix it.
You may change your code to
DATA pp;
SET pp;
if BMI ne . then do:
if BMI
Dear Miriam,
This is SAS bug as mentioned above.
You had better round up the BMI by appropriate digit number.
And please confirm the variable "error" by run the following program, the variable has the smallest error.
data p;
input weight height;
cards;
64 160
;
run;
data pp;
set p;
bmi=weight/(height/100)**2;
error=bmi-25;
run;
data ppp;
set pp;
bmicat=1;
if bmi=. then bmicat=.;
else if 18.5
A different approach is the use of proc format to establish a format, something like that:
proc format;
value bmic 0 - 18.5 = "1"
18.5 - 25 = "0"
25 - 30 = "0.5"
30 - high = "1";
run;
The code must be subbmited only one time, in the SAS section. After, in the data step you can use the format statement to apply the format bmic to the desired variable. The format bmic in the comand receive a dot (.) .
DATA pp;
SET pp;
format bmi bmic.;
The format statement can be used in the proc step instead the data step, but in this case, the relationship between the variable and the format is valid only at the specified proc step.
proc freq;
table bmi;
format bmi bmic.;
run;
Thank you all.
Yuanzhang and Takuto demonstrated nicely how to solve this and I understand much better what happened now.
I can happily continue my analysis :-)
How to create a dummy with first 10 observations of a variables as 1 and rest as 0 in SAS?
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software