Date: Wed, 30 Aug 2000 08:35:33 EDT
Reply-To: David Wright <david_20000@HOTMAIL.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: David Wright <david_20000@HOTMAIL.COM>
Subject: Strange Phenomonon of Running Macro
Content-Type: text/plain; format=flowed
Hi, My Dear friends,
I am so sorry I have to ask help for the followed program again. It run
smoothly but I found there are very strange phenomona among them.
In this program, I want to a REWEIGHT with different RSTU (studentized
residuals). When RSTU is less than 0, I want SWLSL is run since the big
weight will be given to those with less value of RSTU. When RSTU is larger
than zero, I want to run SWLSG since the big weight will be given those with
big RSTU. Just as the followed program shown, I use a switch to realize the
above function.
However, when I test the program by the critical parameters by K=5 %to 6
(when k=5, &&irs&k=-1.26 and &&ire&k=-1.25; when k=6, &&irs&k=1.31 and
&&ire&k=1.32), I found the running of the program is not as what I expected.
The experimental records are as followed:
Program Intercept Xcoefficient Status
RSTU<0 %then %SWLSL 114.3 18 Unexpected
%ELSE %SWLSG 112.4 28 Expected
RSTU<0 %then %SWLSG 114.3 18 Unexpected
%ELSE %SWLSG 110 28.48 Expected
RSTU<0 %then %SWLSG 176.1 -11 Expected
%ELSE %SWLSL 160.4 -1.34 Unexpected
RSTU<0 %then %SWLSL 176.1 -11 Expected
%ELSE %SWLSL 160.4 -1.34 Unexpected
From the above table, the portfolio I expected, which should be correct
calculation and is shown as followed), did not come out.
RSTU<0 %then %SWLSG 176.1 -11 Expected
%ELSE %SWLSL 112.4 28 Expected
It seemed that when RSTU<0, either you run SWLSG or SWLSL, the results are
the same. I can't understand. To testify this hypothesis, I also performed
the experiment that do not have SWITCH. The experimental records are:
K=5 %to 5 (&&irs&k=-1.26 and &&ire&k=-1.25)
Subprogram Intercept Xcoefficient Status
SWLSL 114 18 Unexpected
SWLSG 114 18 Unexpected
K=6 %to 6 (&&irs&k=1.31 and &&ire&k=1.32)
Subprogram Intercept Xcoefficient Status
SWLSL 160 -1.34 Unexpected
SWLSG 112 28.20 Expected
Obviously, in the Case of (K=5 %to 5), either performing SWLSL or SWLSG,
both the results are wrong. Maybe something wrong happened either SWLSL and
SWLSG. But I can't find it after trying one day. I really you give me hands.
I guess it might be some obvious fact or concept I ignore.
Thanks a lot.
David
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
NOTE: Copyright (c) 1989-1996 by SAS Institute Inc., Cary, NC, USA.
NOTE: SAS (r) Proprietary Software Release 6.12 TS020
Licensed to SYRACUSE UNIVERSITY, Site 0009095001.
1
2 %macro Rdata;
3 data sample;
4 infile
"p:\ctip-research\simulation_datasets\high_variance\test&i..txt";
5 input X y;
6 run;
7 %mend Rdata;
8
9 %macro SWLSL(IN=, X=X, Y=Y, RSTU=);
10 filename myfile
'p:\ctip-research\wanted_high\swat_high_1585_error1_runresults\swat_high_1585
_error1_runresults_log.txt';
11 proc printto log=myfile;run;
12 filename myoutput
'p:\ctip-research\wanted_high\swat_high_1585_error1_runresults\swat_high_1585
_error1_runresults_output.txt';
13 proc printto print=myoutput;run;
14 proc reg data=sample;
15 model y=x;
16 reweight student. le &RSTU/weight=1 NOLIST;
17 reweight student. ge &RSTU/weight=0.05 NOLIST;
18 run;
19 PROC PRINTTO;run;
20 PROC PRINTTO;run;
21 %mend SWLSL;
22
23
24 %macro SWLSG(IN=, X=X, Y=Y, RSTU=);
25 filename myfile
'p:\ctip-research\wanted_high\swat_high_1585_error1_runresults\swat_high_1585
_error1_runresults_log.txt';
26 proc printto log=myfile;run;
27 filename myoutput
'p:\ctip-research\wanted_high\swat_high_1585_error1_runresults\swat_high_1585
_error1_runresults_output.txt';
28 proc printto print=myoutput;run;
29 proc reg data=sample;
30 model y=x;
31 reweight student. ge &RSTU/weight=1 NOLIST;
32 reweight student. le &RSTU/weight=0.05 NOLIST;
33 run;
34 PROC PRINTTO;run;
35 PROC PRINTTO;run;
36 %mend SWLSG;
37
38 %macro Mcarlo;
39
40 data s_tlp;
41 infile
"p:\ctip-research\wanted_high\swat_high_1585_error1_loopnumofeachDS\swat_high
_1585_error1_num_DS1-100.txt";
42 input tlpv;
43 run;
44
45 data _null_;
46 set s_tlp;
47 call symput('tlp'||left(_n_), tlpv);
48 run;
49
50 %do i=100 %to 100;
51
52 %Rdata;
53
54 data loopsend;
55 infile
"p:\ctip-research\wanted_high\swat_high_1585_error1_split\swat_high_1585_erro
r1_DS&i..txt";
56 input daset tgtNREW CrStart StrtNUM CrEnd EndNUM;
57 crstart=1000*crstart; crend=1000*crend;
58 run;
59
60 data _null_;
61 set loopsend;
62 call symput('irs'||left(_n_), crstart);
63 call symput('ire'||left(_n_), crend);
64 run;
65
66 %do k=1 %to &&tlp&i;
67
68 %do;
69 %do j=&&irs&k %to &&ire&k;
70
71 %let RSTU= %sysevalf(0.001*(&j));
72 filename myfile '
73
p:\ctip-research\wanted_high\swat_high_1585_error1_runresults\swat_high_1585_
error1_runresults_log.txt';
74 proc printto log=myfile;run;
75 %put The Studentized Residual for Comparing is:
&RSTU;
76 PROC PRINTTO;run;
77
78 %if rstu <0 %then
79 %do;
80 %SWLSG(IN=sample, X=X, Y=Y, RSTU=&RSTU);
81 %end;
82 %else
83 %do;
84 %SWLSL(IN=sample, X=X, Y=Y, RSTU=&RSTU);
85 %end;
86 %end;
87 %end;
88 %end;
89
90 %end;
91
92 %mend Mcarlo;
93
94 %Mcarlo;
_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
Share information about yourself, create your own public profile at
http://profiles.msn.com.
