how to use proc iml to log transform negative values.

[Cruise]

Hi Folks:

I've never used proc iml before and trying to use the approach suggested in the Wicklin's blog. 

I have a dataset with a variable with negative values which needs to be log transformed. 

data have;
input var;
cards;
-1
-2
; 

And this step was suggested. Can you help me understand how I am supposed use the proc iml approach suggested so that variable 'var' with negative values in my 'have' dataset is translated and log transformed? 

proc iml;
Y = {-3,1,2,.,5,10,100};    /* negative datum */
LY = log10(Y + 1 - min(Y)); /* translate, then transform */

https://blogs.sas.com/content/iml/2011/04/27/log-transformations-how-to-handle-negative-data-values....

[Reeza · Accepted solution]

Y creates the vector there. You have a vector of -1, -2. You need to first load your data from a SAS data set into IML as a vector or matrix, since you only have one column a vector is fine. Then you can jump to the LY portion.

This blog post has an example.
https://blogs.sas.com/content/iml/2012/05/14/how-to-read-data-set-variables-into-sasiml-vectors.html

[Cruise]

@Reeza, thanks a lot. Trying to understand the example. My actual data has about 1600 rows and 5 variables that all needed to be log transformed.

[Cruise]

@Reeza 

Reeza, I get the following error. Also, after transforming the variable min_min_temp, then, how to integrate this vector back to my data set run my model on? I'm dumb founded. Is it possible that you showcase on the 'have' data, please ? 

data have; 
input date	min_min_temp max_max_temp;
cards;
20454	-5.9	12
20455	-2.7	16.8
20456	-2.3	16.6
20457	-3.2	15.2
20458	-6.1	11.1
20459	-7.8	9.9
20460	-7.6	8.5
20461	-8.9	7.5
20462	-6.2	9.4
20463	-3.9	10.9
;

5296  proc iml;
NOTE: IML Ready
5297  use weather1;
5298  read var {min_min_temp};
5299  Y = {min_min_temp};
5299!                        /* negative datum */
5300  Log_min_temp = log(Y + 1 - min(Y));
ERROR: (execution) Numeric argument should be character.

 operation : + at line 5300 column 22
 operands  : Y, *LIT1003

Y      1 row       1 col     (character, size 12)

 MIN_MIN_TEMP

*LIT1003      1 row       1 col     (numeric)

         1

 statement : ASSIGN at line 5300 column 1
5300!

[Reeza]

Why are you using IML then? Why not keep everything in a data step?

[Cruise]

Oops, yes, I should indeed 🙂

[Reeza]

1. Run a proc means to get the minimum for each variable you want to analyze

2. Merge that in to your data

3. Set up two arrays, one for the raw data and one for the transformed data

4. Set up an array for the means (temporary is also good, but not really needed)

4. Loop through and do your transformations

Adding summary statistics to a data set

https://github.com/statgeek/SAS-Tutorials/blob/master/add_average_value_to_dataset.sas

Here's a tutorial on using Arrays in SAS
https://stats.idre.ucla.edu/sas/seminars/sas-arrays/

And an entirely different suggestion - look into the options in PROC STDIZE for data transformations but none seem close to what you want to do. Just something to keep in mind though.