topic SAS Function for Curve Fitting in SAS Enterprise Guide

SAS Function for Curve Fitting

siva5578 — Wed, 06 Nov 2019 15:44:10 GMT

Please refer data below

	Customer1	Customer2	Customer3	Customer4	Customer5
Jan-18	$5,000	$8,000	$5,250	$6,125	$3,000
Feb-18	$5,100	$8,300	$5,500	$6,625	$3,050
Mar-18	$5,200	$8,600	$5,750	$7,125	$3,100
Apr-18	$5,300	$8,900	$6,000	$7,625	$3,150
May-18	$5,400	$9,200	$6,250	$8,125	$3,200
Jun-18	$5,500	$9,500	$6,500	$8,625	$3,250
Jul-18	$5,600	$9,800	$6,750	$9,125	$3,300
Aug-18	$5,700	$10,100	$7,000	$9,625	$3,350
Sep-18	$5,800	$10,400	$7,250	$10,125	$3,400
Oct-18	$5,900	$10,700	$7,500	$10,625	$3,450
Nov-18	$6,000	$11,000	$7,750	$11,125	$3,500
Dec-18	$6,100	$11,300	$8,000	$11,625	$3,550

For each customer, I am trying to create a curve that fits the past 12 months of data. Essentially, if it is a Polynomial fit (Y = Ax^2+Bx+C ), I am looking for coefficients A,B, C & the error ( MSE or RMSE etc..).

I have used Proc Orthoreg/Reg etc.. to get the results but, this will alow me to do this for 1 account at a time. If I am dealing with large volume of accts, this is not an option at all. So, I don't think I can use SAS procedures for this.

Is/Are there function(s) available in SAS that allows me to do that. I am looking for something similar to the "Polyfit" function in Python.

If not, is there a way to achieve the same results using SAS?

Thanks

Re: SAS Function for Curve Fitting

PaigeMiller — Wed, 06 Nov 2019 15:52:53 GMT

You don't really indicate which variable is X and which variable is Y is in this data, so could you please clarify this?

In general, a polynomial function is fit through PROC GLM (although there are many other PROCs that will do this as well). Example:

proc glm data=have;
    model y = x x*x;
run;
quit;

Re: SAS Function for Curve Fitting

PeterClemmensen — Wed, 06 Nov 2019 15:49:27 GMT

What is your dependent variable here?

Re: SAS Function for Curve Fitting

siva5578 — Wed, 06 Nov 2019 16:04:19 GMT

The independent variable would be time period (month). The dependent variable would be the values listed in columns for each customer.

As I mentioned, I am trying to independently curve fit for each customer.

Re: SAS Function for Curve Fitting

PaigeMiller — Wed, 06 Nov 2019 16:10:09 GMT

I added example code to my first reply. That should give you enough information to do this.

Re: SAS Function for Curve Fitting

siva5578 — Wed, 06 Nov 2019 17:11:41 GMT

Paige,

Thanks for the response.

I have used Proc Orthoreg/Reg etc.. to get the results (Coefficients for the Polynomial equation), 1 account at a time. I think PROC GLM would work the same way. In the sample table I had provided, I would have to create separate SAS dataset for each customer and use PROC GLM for each of those datasets. If I am dealing with large volume of accts, this is not an option at all. So, I don't think I can use SAS PROC for this. Please correct me if I misunderstood how Proc GLM would work.

I think a SAS Function would be the way to go. Let me know your thoughts.

Thanks.

Siva

Re: SAS Function for Curve Fitting

PaigeMiller — Wed, 06 Nov 2019 17:41:51 GMT

@siva5578 wrote:

In the sample table I had provided, I would have to create separate SAS dataset for each customer and use PROC GLM for each of those datasets.

No, one data set works fine.

If I am dealing with large volume of accts, this is not an option at all. So, I don't think I can use SAS PROC for this. Please correct me if I misunderstood how Proc GLM would work.

Sure it works if you follow these instructions: https://blogs.sas.com/content/iml/2017/02/13/run-1000-regressions.html