Date: Thu, 17 Dec 2009 18:53:00 -0800
Reply-To: XL <wangxiaolu@YAHOO.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: XL <wangxiaolu@YAHOO.COM>
Organization: http://groups.google.com
Subject: Re: proc loess problem
Content-Type: text/plain; charset=ISO-8859-1
On Dec 17, 11:06 am, HERMA...@WESTAT.COM (Sigurd Hermansen) wrote:
> How many distinct values does the class "state" have? The BY statement essentially calls for that many different parameter estimate based on state subsets of data. You may have to reconsider the role of the state variable in the model.
>
> The iterations option could also require extra memory. Do the observed errors have a normal distribution as assumed?
> S
>
>
>
> -----Original Message-----
> From: SAS(r) Discussion [mailto:SA...@LISTSERV.UGA.EDU] On Behalf Of XL
> Sent: Wednesday, December 16, 2009 11:54 PM
> To: SA...@LISTSERV.UGA.EDU
> Subject: proc loess problem
>
> I am trying to smooth fit for my dataset. My dataset contains about
> 30,000 data points.
>
> I either get 'out of memory', or SAS just stops working. Please see my
> code below.The problem occurs when I add 'clm' and 'r' option. I know
> it will significantly increase the computation. But I need the CLM.
>
> proc loess data=sample;
> by state;
> model y = x / direct clm r iterations=3;
> ods output modelsummary=modelsum outputstatistics=modelout;
> run;
>
> I try to increase memsize to 4G. But it doesnt help. SAS stops working
> even before used RAM reaching 4G.
>
> Anybody can help? Thanks.
>
> By the way, I am using SAS 9.2 on Windows 7.- Hide quoted text -
>
> - Show quoted text -
Thanks for your reply Sigurd,
State is a dummy variable, taking either 0 or 1. And the errors do not
seem to be normal.
Is there any way that can let SAS literally use the RAM available? My
computer has 3G RAM, but SAS stops before using up all the RAM.
I tried memsize= option, but doesnt seem to help. Any other way that I
might try?
Xiaolu
|