一个关于proc reg的问题

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

[size=1em]calculating PRESSp (prediction sum of squares) criterion

Hi all,

As I'm doing a regression in which I want to reduce my current amount of possible parameters, I wanted to run a kind of best-subset algorithm in that I want to evaluate (and plot) each possible combination of regression parameters using some model selection criteria as R2 (adjusted), MSE, Cp, AICp, SBCp and PRESSp.

As such, I am now able to generate a listing of all possible regression parameter combinations and evaluating them against R2, R2 adjusted, Cp, MSE and SSE using the following command:

proc reg data=model2 outest=temp;

model lny = X3-X8 X12-X23 X4X16 /selection=rsquare adjrsq cp press mse sse;

plot cp.*np.;

plot rsq.*np.;

plot mse.*np.;

run;

quit;

However, I would really like to have the PRESSp (preidction sum of squares) value calculated (which to former command apparently doesn't do?) and plotted and, if possible, also Akaike's information criterion (AICp) and Schwarz' Bayesian criterion (SBCp).

Could anyone please help?

Kind regards,

Peter

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：REG ROC Combinations combination Calculating prediction adjusted criteria generate possible

许多[size=1em]是什么意思?

楼主是复制、粘贴的吧

goldbaodi 发表于 2012-4-2 09:55
许多是什么意思?

恩，复制的时候带了些网页代码。这个问题，是别人提出来的，我刚好也不会，所以看看有没有人能解决。

gavinking 发表于 2012-4-3 03:19
恩，复制的时候带了些网页代码。这个问题，是别人提出来的，我刚好也不会，所以看看有没有人能解决。

statistic PRESS is NOT available for selection = Rsquare.
JingJu