I want to normalize my data using log10 in SAS. Please write the related program for me.
Thank you
Can you clarify what you mean by "normalize" your data? Do you mean:
1) Log-transform a variable so that a non-normally distributed variable now takes a normal distribution
2) "Center" a variable (so that the values are presented as z-scores)
3) Something else that I haven't thought of...
Hi all,
I did normality test in SAS program. the results showed that data are not normal. So, i have to transform my data with log10. in fact, i need this program (transform data with log10). Thank you for your answer
Thank you for the clarification. So you are asking how to transform one variable that currently has a non-normal distribution using a logarithmic transformation in the hopes that it will give an approximately normal distribution. The necessary SAS code is as follows:
DATA dataset2;
SET dataset1;
LOGVAR=log10(VAR);
RUN;
This would create a new variable LOGVAR that is the log10 transformation of the variable VAR.
Dear Mohammadreza
Any logaritmic transformation produces the same efect on the normality of a variable. Because Log10(x)= logb (X) log10(b), where b is any logaritmic base, and log10(b) is a known constant.
Then in the program of Andrew, you may use the natural logaritmic or any other.
For the natural logaritmic it would be:
DATA dataset2;
SET dataset1;
LOGVAR=log(VAR);
RUN;
I do not make comment on how you could perform log10 transformation in SAS.
Only one minor comment: In case what you expect (log10 transformation) could not bring your data into normality, then you might need to think to other transformations as well (if it is meaningful to do so). That is when you might think to use Box-Cox transformation (I recommend you read section 4.8 (Transformation to near normality) in the Book "Applied Multivariate Statistical Analysis" by Johnson and Wichern).
Hi Dear Mohammadreza Pourghayoumi,
I write a formula to you, I hope it will be useful:
data Mehdi;
input trt rep IC;
IC1=log(IC);
cards;
1 1 21.1
1 2 -3
1 3 4
1 4 -0.79
1 5 -3.47
1 6 -20.99
2 1 1.19
2 2 7.7
2 3 -22.93
2 4 -21.7
2 5 0.26
2 6 8.9
3 1 -23.32
3 2 -24.97
3 3 -2.63
3 4 -2.36
3 5 -11.86
3 6 -1.27
4 1 -2.85
4 2 .
4 3 1.1
4 4 2.62
4 5 5.28
4 6 3.87
;
/*proc means mean std stderr cv min max;*/
proc means mean stderr ;
class trt;
var IC;
run;
proc glm;
class trt rep;
model IC= trt;
output out = omatti1 residual = omatti2;
run;
proc capability data = omatti1 normal;
var omatti2;
run;
Regards,
Mehdi
Dear Mohammad Mehdi,
Thank you very much for your useful help.
Wishing you the best,
Dear Mohammadreza and Mohammad
I notice that in the example listed above, several values are negative. If you use it the logaritms good be invalid, and replaced by missing. If any value were 0 it would gave the same effect.
Then if the data above were the actual data, to normalice it ,requires to add a positive value greater than the smaller negative value. For example adding 25 to each value;
The proc glm and means should be applied to IC1 instead of IC.
Then:
data Mehdi;
input trt rep IC;
ric=ic+25;
IC1=log(RIC);
cards;
1 1 21.1
1 2 -3
1 3 4
1 4 -0.79
1 5 -3.47
1 6 -20.99
2 1 1.19
2 2 7.7
2 3 -22.93
2 4 -21.7
2 5 0.26
2 6 8.9
3 1 -23.32
3 2 -24.97
3 3 -2.63
3 4 -2.36
3 5 -11.86
3 6 -1.27
4 1 -2.85
4 2 .
4 3 1.1
4 4 2.62
4 5 5.28
4 6 3.87
;
proc means mean stderr ;
class trt;
var IC1;
run;
proc glm;
class trt rep;
model IC1= trt;
output out = omatti1 residual = omatti2;
run;
proc capability data = omatti1 normal;
var omatti2;
run;
Dear Dr Guillermo,
Thank you for your good suggestion about this important subject.
Best regards,
Dear DR Guillermo Enrique Ramos ,
Thank you so much for your insight, but as you refer, I wanted to write only this proc for him, but you are right and I appreciate your helping.
Regards,
Mehid