Calcite | Level 5

## Run linear model on computed orthogonal polynomial points to the 7th degree and display predictions.

Hi,

I am trying to run a simple regression in base SAS that uses the orthogonal polynomial to the 7th degree and then append predictions to my data set. I do not have SAS IML so I cannot use this function. In R the code would simply be poly_model = lm(y ~ poly(x,n)) where

y= { 8.249098 7.621568 7.199557 8.371507 7.927080 7.691783 7.662529 7.170472 7.562522 7.889212 8.139248
7.695632 7.502941 7.914134 7.201237 7.963647 7.860750}

x={2001 2008 2016 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016}

n=7

This is how my data is structured in SAS:

data Model_data;

input Year Factor1 Factor2;

datalines;

2001    8.249098   .

2003    .                 8.371507

2004    .                 7.927080

2005    .                 7.691783

2006    .                 7.662529

2007    .                 7.170472

2008    7.621568   7.562522

2009    .                 7.889212

2010    .                 8.139248

2011    .                 7.695632

2012    .                 7.502941

2013    .                 7.914134

2014    .                 7.201237

2015    .                 7.963647

2016    7.199557   7.860750;

run;

I would like the output to be a datatable with variables Year Factor1 Factor2 Predictions.

I have tried running:

proc orthoreg data=model_data;
effect xMod = poly(Year1/ degree=7 details standardize(method= wmoments) standardize=center);
model Y = xMod;
run;

However, this did not yield the same results as the R code lm(y~poly(x,n)). I suspect this is because SAS centers and normalizes the data differently.

Any hints or help with this would be greatly appreciated.

SAS Super FREQ

## Re: Run linear model on computed orthogonal polynomial points to the 7th degree and display predicti

I didn't check the options that you are using on the MODEL statement, but here is the DATA step that you should use to generate the data:

``````data Model_Data;
input x y;
datalines;
2001 8.249098
2008 7.621568
2016 7.199557
2003 8.371507
2004 7.92708
2005 7.691783
2006 7.662529
2007 7.170472
2008 7.562522
2009 7.889212
2010 8.139248
2011 7.695632
2012 7.502941
2013 7.914134
2014 7.201237
2015 7.963647
2016 7.86075
;

``````

For more about polynomials effects and the EFFECT statement, see "Construct polynomial effects in SAS regression models."

Discussion stats