problem in glm and glmselect (revised)

[jhmoon]

Dear all,

I posted same title 2 weeks ago, and I do it again after making it clear to see.

previous post: https://communities.sas.com/t5/Statistical-Procedures/Different-results-between-glm-and-glmselect/td...

My problem is that when I did a  event study analysis with code 'glm' and 'glmselect', the results are not the same, and I want to know why.

Specifically, I use event-study method, and the outcome variable is "calc_cpi", RHS variables are 9 event time dummies, 18 year dummies and 18 age dummies. 

Weight variable here is constructed by this way: the number of cohort 2 / the number of cohort 1 (by age at event occurs).

So, cohort 2 has 1 for the weight for all samples. Cohort 1 has the weight by age at event time zero.

Case 1: all the results are right both by glm and glmselect without weights applied.

Case 2: 'glmselect' has a problem. Here's the code and output. (first one is glm result, and second one is glmselect result)

PROC GLM DATA=IPW.TREAT_BASE_SAMPLE_CHT70  ;
CLASS EVENT_TIME(REF='-1') AGE(REF='39') STD_YYYY(REF='2019');
MODEL CALC_CPI = EVENT_TIME AGE STD_YYYY/ SOLUTION;
WEIGHT C1_WEIGHT;
OUTPUT OUT= PRED_CALC_CPI_WC1 PREDICTED=P RESIDUAL=R;
ODS OUTPUT ParameterEstimates=PARAM_CALC_CPI_WC1;
RUN;


PROC GLMSELECT DATA=IPW.TREAT_BASE_SAMPLE_CHT70 ;
CLASS EVENT_TIME(REF='-1') AGE STD_YYYY ;
MODEL CALC_CPI = EVENT_TIME AGE STD_YYYY/ SELECTION=NONE ;
WEIGHT C1_WEIGHT;
OUTPUT OUT= PRED_CALC_CPI_WC1 PREDICTED=P RESIDUAL=R;
ODS OUTPUT ParameterEstimates=PARAM_CALC_CPI_WC1;
RUN;

Dependent	Parameter	Estimate	Biased	StdErr	tValue	Probt
CALC_CPI	Intercept	57493.58084	1	598.5360913	96.06	<.0001
CALC_CPI	EVENT_TIME -3	1151.43048	1	67.746409	17	<.0001
CALC_CPI	EVENT_TIME -2	1290.48506	1	63.7409616	20.25	<.0001
CALC_CPI	EVENT_TIME 0	-6709.90206	1	63.2582534	-106.07	<.0001
CALC_CPI	EVENT_TIME 1	-16152.14249	1	65.8422001	-245.32	<.0001
CALC_CPI	EVENT_TIME 2	-18193.03078	1	67.8515907	-268.13	<.0001
CALC_CPI	EVENT_TIME 3	-19273.77713	1	70.6162547	-272.94	<.0001
CALC_CPI	EVENT_TIME 4	-20520.90805	1	73.7837183	-278.12	<.0001
CALC_CPI	EVENT_TIME 5	-21161.09299	1	77.6282235	-272.6	<.0001
CALC_CPI	EVENT_TIME -1	0	1
CALC_CPI	AGE        22	-33011.33226	1	453.9691843	-72.72	<.0001
CALC_CPI	AGE        23	-31030.13856	1	417.0497166	-74.4	<.0001
CALC_CPI	AGE        24	-28456.20384	1	404.3707065	-70.37	<.0001
CALC_CPI	AGE        25	-26188.46721	1	398.0201918	-65.8	<.0001
CALC_CPI	AGE        26	-24184.5052	1	394.2489703	-61.34	<.0001
CALC_CPI	AGE        27	-22544.71582	1	391.5407826	-57.58	<.0001
CALC_CPI	AGE        28	-20991.74813	1	389.4779456	-53.9	<.0001
CALC_CPI	AGE        29	-19253.74307	1	387.720454	-49.66	<.0001
CALC_CPI	AGE        30	-17464.85105	1	386.0938409	-45.23	<.0001
CALC_CPI	AGE        31	-15496.64241	1	384.8382356	-40.27	<.0001
CALC_CPI	AGE        32	-13496.68391	1	383.4668709	-35.2	<.0001
CALC_CPI	AGE        33	-11624.09823	1	381.9453759	-30.43	<.0001
CALC_CPI	AGE        34	-9663.06563	1	380.3173421	-25.41	<.0001
CALC_CPI	AGE        35	-7642.13793	1	377.3506788	-20.25	<.0001
CALC_CPI	AGE        36	-5740.0934	1	374.7912852	-15.32	<.0001
CALC_CPI	AGE        37	-3900.21078	1	373.8001484	-10.43	<.0001
CALC_CPI	AGE        38	-1949.43689	1	377.7374853	-5.16	<.0001
CALC_CPI	AGE        39	0	1
CALC_CPI	STD_YYYY   2002	-16249.47242	1	723.7739723	-22.45	<.0001
CALC_CPI	STD_YYYY   2003	-13821.10908	1	715.5893296	-19.31	<.0001
CALC_CPI	STD_YYYY   2004	-11888.8523	1	712.2320853	-16.69	<.0001
CALC_CPI	STD_YYYY   2005	-11398.111	1	710.29501	-16.05	<.0001
CALC_CPI	STD_YYYY   2006	-9609.67612	1	708.9337932	-13.56	<.0001
CALC_CPI	STD_YYYY   2007	-7473.76668	1	707.8289777	-10.56	<.0001
CALC_CPI	STD_YYYY   2008	-5956.85424	1	706.8830577	-8.43	<.0001
CALC_CPI	STD_YYYY   2009	-5790.1094	1	706.0041004	-8.2	<.0001
CALC_CPI	STD_YYYY   2010	-5811.82223	1	705.1423241	-8.24	<.0001
CALC_CPI	STD_YYYY   2011	-5341.59143	1	704.5219159	-7.58	<.0001
CALC_CPI	STD_YYYY   2012	-4584.69526	1	703.8424319	-6.51	<.0001
CALC_CPI	STD_YYYY   2013	-4165.1198	1	703.2508081	-5.92	<.0001
CALC_CPI	STD_YYYY   2014	-3306.6357	1	702.6738477	-4.71	<.0001
CALC_CPI	STD_YYYY   2015	-2796.68722	1	700.2987691	-3.99	<.0001
CALC_CPI	STD_YYYY   2016	-2641.5268	1	700.5926244	-3.77	0.0002
CALC_CPI	STD_YYYY   2017	-2145.88724	1	704.8095795	-3.04	0.0023
CALC_CPI	STD_YYYY   2018	-1733.12006	1	722.7610169	-2.4	0.0165
CALC_CPI	STD_YYYY   2019	0	1

Effect	EVENT_TIME	AGE	STD_YYYY	Parameter	DF	Estimate	StandardizedEst	StdErr	tValue	Probt
Intercept				Intercept	1	5616.721735	0	5861231.563	0.00	0.9992
EVENT_TIME	-3			EVENT_TIME -3	1	1151.430482	0.010214407	67.74641492	17.00	0.0000
EVENT_TIME	-2			EVENT_TIME -2	1	1290.485072	0.011445652	63.74096714	20.25	0.0000
EVENT_TIME	0			EVENT_TIME 0	1	-6709.902053	-0.059401827	63.25825897	-106.07	0.0000
EVENT_TIME	1			EVENT_TIME 1	1	-16152.14248	-0.139525561	65.84220583	-245.32	0.0000
EVENT_TIME	2			EVENT_TIME 2	1	-18193.03077	-0.159262176	67.85159667	-268.13	0.0000
EVENT_TIME	3			EVENT_TIME 3	1	-19273.77712	-0.16878725	70.61626087	-272.94	0.0000
EVENT_TIME	4			EVENT_TIME 4	1	-20520.90803	-0.179322043	73.78372473	-278.12	0.0000
EVENT_TIME	5			EVENT_TIME 5	1	-21161.09296	-0.185167699	77.62823033	-272.60	0.0000
EVENT_TIME	-1			EVENT_TIME -1	0	0	0
AGE		22		AGE        22	1	-33011.33219	-0.072258126	453.9692229	-72.72	0.0000
AGE		23		AGE        23	1	-31030.13848	-0.105624115	417.0497519	-74.40	0.0000
AGE		24		AGE        24	1	-28456.20376	-0.129083905	404.3707406	-70.37	0.0000
AGE		25		AGE        25	1	-26188.46713	-0.148142386	398.0202253	-65.80	0.0000
AGE		26		AGE        26	1	-24184.50513	-0.16058298	394.2490035	-61.34	0.0000
AGE		27		AGE        27	1	-22544.71575	-0.169706244	391.5408156	-57.58	0.0000
AGE		28		AGE        28	1	-20991.74806	-0.171072624	389.4779784	-53.90	0.0000
AGE		29		AGE        29	1	-19253.743	-0.164343939	387.7204866	-49.66	0.0000
AGE		30		AGE        30	1	-17464.85098	-0.152418958	386.0938734	-45.23	0.0000
AGE		31		AGE        31	1	-15496.64234	-0.132617654	384.8382679	-40.27	0.0000
AGE		32		AGE        32	1	-13496.68384	-0.11066162	383.4669031	-35.20	0.0000
AGE		33		AGE        33	1	-11624.09816	-0.089483332	381.9454081	-30.43	0.0000
AGE		34		AGE        34	1	-9663.065559	-0.067725564	380.3173741	-25.41	0.0000
AGE		35		AGE        35	1	-7642.137868	-0.046998557	377.3507106	-20.25	0.0000
AGE		36		AGE        36	1	-5740.093333	-0.028803712	374.7913168	-15.32	0.0000
AGE		37		AGE        37	1	-3900.210716	-0.01479889	373.8001801	-10.43	0.0000
AGE		38		AGE        38	1	-1949.436833	-0.005011604	377.7375174	-5.16	0.0000
AGE		39		AGE        39	0	0	0
STD_YYYY			2002	STD_YYYY   2002	1	35627.3866	0.142964368	5861231.573	0.01	0.9952
STD_YYYY			2003	STD_YYYY   2003	1	38055.74994	0.207465875	5861231.572	0.01	0.9948
STD_YYYY			2004	STD_YYYY   2004	1	39988.00673	0.262688821	5861231.571	0.01	0.9946
STD_YYYY			2005	STD_YYYY   2005	1	40478.74803	0.297424003	5861231.571	0.01	0.9945
STD_YYYY			2006	STD_YYYY   2006	1	42267.1829	0.333966531	5861231.571	0.01	0.9942
STD_YYYY			2007	STD_YYYY   2007	1	44403.09234	0.369594865	5861231.571	0.01	0.9940
STD_YYYY			2008	STD_YYYY   2008	1	45920.00478	0.393761508	5861231.571	0.01	0.9937
STD_YYYY			2009	STD_YYYY   2009	1	46086.74962	0.401552632	5861231.571	0.01	0.9937
STD_YYYY			2010	STD_YYYY   2010	1	46065.03678	0.402040126	5861231.571	0.01	0.9937
STD_YYYY			2011	STD_YYYY   2011	1	46535.26758	0.370095463	5861231.571	0.01	0.9937
STD_YYYY			2012	STD_YYYY   2012	1	47292.16376	0.338098445	5861231.571	0.01	0.9936
STD_YYYY			2013	STD_YYYY   2013	1	47711.73922	0.296078403	5861231.571	0.01	0.9935
STD_YYYY			2014	STD_YYYY   2014	1	48570.22332	0.256641829	5861231.571	0.01	0.9934
STD_YYYY			2015	STD_YYYY   2015	1	49080.1718	0.212832555	5861231.571	0.01	0.9933
STD_YYYY			2016	STD_YYYY   2016	1	49235.33223	0.162905538	5861231.571	0.01	0.9933
STD_YYYY			2017	STD_YYYY   2017	1	49730.97179	0.116719528	5861231.572	0.01	0.9932
STD_YYYY			2018	STD_YYYY   2018	1	50143.73898	0.071874329	5861231.575	0.01	0.9932
STD_YYYY			2019	STD_YYYY   2019	1	51876.85768	0.035889768	5861231.593	0.01	0.9929

With glmselect code, STD_YYYY has no reference dummy which should be omitted.

Case 3: When I do a subgroup analysis with 'by' option, there's a different problem here.

'Metro' variable is a dummy for indicating the metro city.

PROC GLM DATA=IPW.TREAT_BASE_SAMPLE_CHT70  ;
CLASS EVENT_TIME(REF='-1') AGE(REF='39') STD_YYYY(REF='2019');
MODEL CALC_CPI = EVENT_TIME AGE STD_YYYY/ SOLUTION;
BY METRO;
WEIGHT C1_WEIGHT;
OUTPUT OUT= PRED_CALC_CPI_WC1 PREDICTED=P RESIDUAL=R;
ODS OUTPUT ParameterEstimates=PARAM_CALC_CPI_WC1;
RUN;

PROC GLMSELECT DATA=IPW.TREAT_BASE_SAMPLE_CHT70 ;
CLASS EVENT_TIME(REF='-1') AGE STD_YYYY ;
MODEL CALC_CPI = EVENT_TIME AGE STD_YYYY/ SELECTION=NONE ;
BY METRO;
WEIGHT C1_WEIGHT;
OUTPUT OUT= PRED_CALC_CPI_WC1 PREDICTED=P RESIDUAL=R;
ODS OUTPUT ParameterEstimates=PARAM_CALC_CPI_WC1;
RUN;

METRO	Dependent	Parameter	Estimate	Biased	StdErr	tValue	Probt
0	CALC_CPI	Intercept	-3895.12278	1	7729975.654	0.00	0.9996
0	CALC_CPI	EVENT_TIME -3	978.61218	1	87.79501768	11.15	0.0000
0	CALC_CPI	EVENT_TIME -2	1180.69630	1	82.27551899	14.35	0.0000
0	CALC_CPI	EVENT_TIME 0	-6342.50198	1	81.54173762	-77.78	0.0000
0	CALC_CPI	EVENT_TIME 1	-14450.89523	1	84.74718996	-170.52	0.0000
0	CALC_CPI	EVENT_TIME 2	-16030.37882	1	87.51248946	-183.18	0.0000
0	CALC_CPI	EVENT_TIME 3	-16749.01300	1	91.07836901	-183.90	0.0000
0	CALC_CPI	EVENT_TIME 4	-17470.41779	1	95.08922474	-183.73	0.0000
0	CALC_CPI	EVENT_TIME 5	-17515.59945	1	99.91195499	-175.31	0.0000
0	CALC_CPI	EVENT_TIME -1	0.00000	1
0	CALC_CPI	AGE        22	3726.46625	0	16834818.1	0.00	0.9998
0	CALC_CPI	AGE        23	5682.55616	0	16834818.1	0.00	0.9997
0	CALC_CPI	AGE        24	8018.60850	0	16834818.1	0.00	0.9996
0	CALC_CPI	AGE        25	9747.53333	0	16834818.1	0.00	0.9995
0	CALC_CPI	AGE        26	11121.30219	0	16834818.1	0.00	0.9995
0	CALC_CPI	AGE        27	12108.09981	0	16834818.1	0.00	0.9994
0	CALC_CPI	AGE        28	13026.26507	0	16834818.1	0.00	0.9994
0	CALC_CPI	AGE        29	14138.23528	0	16834818.1	0.00	0.9993
0	CALC_CPI	AGE        30	15416.94606	0	16834818.1	0.00	0.9993
0	CALC_CPI	AGE        31	16866.27534	0	16834818.1	0.00	0.9992
0	CALC_CPI	AGE        32	18346.09867	0	16834818.1	0.00	0.9991
0	CALC_CPI	AGE        33	19629.98543	0	16834818.1	0.00	0.9991
0	CALC_CPI	AGE        34	20940.93670	0	16834818.1	0.00	0.9990
0	CALC_CPI	AGE        35	22268.69419	0	16834818.1	0.00	0.9989
0	CALC_CPI	AGE        36	23498.42169	0	16834818.1	0.00	0.9989
0	CALC_CPI	AGE        37	24639.75805	0	16834818.1	0.00	0.9988
0	CALC_CPI	AGE        38	26288.30791	0	16834818.1	0.00	0.9988
0	CALC_CPI	AGE        39	27677.95218	0	16834818.11	0.00	0.9987
0	CALC_CPI	STD_YYYY   2002	8299.40414	0	16722040.85	0.00	0.9996
0	CALC_CPI	STD_YYYY   2003	10663.93296	0	16722040.85	0.00	0.9995
0	CALC_CPI	STD_YYYY   2004	12542.76508	0	16722040.85	0.00	0.9994
0	CALC_CPI	STD_YYYY   2005	13203.85921	0	16722040.85	0.00	0.9994
0	CALC_CPI	STD_YYYY   2006	14851.59210	0	16722040.85	0.00	0.9993
0	CALC_CPI	STD_YYYY   2007	16799.61320	0	16722040.85	0.00	0.9992
0	CALC_CPI	STD_YYYY   2008	18178.73109	0	16722040.85	0.00	0.9991
0	CALC_CPI	STD_YYYY   2009	18532.44842	0	16722040.85	0.00	0.9991
0	CALC_CPI	STD_YYYY   2010	18300.38044	0	16722040.85	0.00	0.9991
0	CALC_CPI	STD_YYYY   2011	18846.57380	0	16722040.85	0.00	0.9991
0	CALC_CPI	STD_YYYY   2012	19583.46699	0	16722040.85	0.00	0.9991
0	CALC_CPI	STD_YYYY   2013	20241.00004	0	16722040.85	0.00	0.9990
0	CALC_CPI	STD_YYYY   2014	21323.54965	0	16722040.85	0.00	0.9990
0	CALC_CPI	STD_YYYY   2015	21817.24317	0	16722040.85	0.00	0.9990
0	CALC_CPI	STD_YYYY   2016	22159.80750	0	16722040.85	0.00	0.9989
0	CALC_CPI	STD_YYYY   2017	22649.12430	0	16722040.86	0.00	0.9989
0	CALC_CPI	STD_YYYY   2018	23302.41868	0	16722040.86	0.00	0.9989
0	CALC_CPI	STD_YYYY   2019	24949.93483	0	16722040.88	0.00	0.9988
1	CALC_CPI	Intercept	42176.43379	1	8924444.166	0.00	0.9962
1	CALC_CPI	EVENT_TIME -3	994.34186	1	101.1486969	9.83	0.0000
1	CALC_CPI	EVENT_TIME -2	1215.89427	1	95.45162941	12.74	0.0000
1	CALC_CPI	EVENT_TIME 0	-6843.48301	1	94.85321721	-72.15	0.0000
1	CALC_CPI	EVENT_TIME 1	-17267.81526	1	98.94975143	-174.51	0.0000
1	CALC_CPI	EVENT_TIME 2	-19545.15122	1	101.8492618	-191.90	0.0000
1	CALC_CPI	EVENT_TIME 3	-20748.45919	1	106.0721071	-195.61	0.0000
1	CALC_CPI	EVENT_TIME 4	-22258.63616	1	110.9774742	-200.57	0.0000
1	CALC_CPI	EVENT_TIME 5	-23210.70218	1	116.9501239	-198.47	0.0000
1	CALC_CPI	EVENT_TIME -1	0.00000	1
1	CALC_CPI	AGE        22	-16614.47689	1	8924444.227	0.00	0.9985
1	CALC_CPI	AGE        23	-14644.87904	1	8924444.222	0.00	0.9987
1	CALC_CPI	AGE        24	-11836.91128	1	8924444.221	0.00	0.9989
1	CALC_CPI	AGE        25	-9061.92505	1	8924444.22	0.00	0.9992
1	CALC_CPI	AGE        26	-6561.32080	1	8924444.22	0.00	0.9994
1	CALC_CPI	AGE        27	-4455.03090	1	8924444.219	0.00	0.9996
1	CALC_CPI	AGE        28	-2508.50369	1	8924444.219	0.00	0.9998
1	CALC_CPI	AGE        29	-433.24748	1	8924444.219	0.00	1.0000
1	CALC_CPI	AGE        30	1583.46575	1	8924444.219	0.00	0.9999
1	CALC_CPI	AGE        31	3797.79359	1	8924444.219	0.00	0.9997
1	CALC_CPI	AGE        32	6034.52033	1	8924444.219	0.00	0.9995
1	CALC_CPI	AGE        33	8181.42850	1	8924444.219	0.00	0.9993
1	CALC_CPI	AGE        34	10412.49360	1	8924444.219	0.00	0.9991
1	CALC_CPI	AGE        35	12734.45466	1	8924444.218	0.00	0.9989
1	CALC_CPI	AGE        36	14889.39073	1	8924444.218	0.00	0.9987
1	CALC_CPI	AGE        37	17002.81232	1	8924444.218	0.00	0.9985
1	CALC_CPI	AGE        38	18969.41217	1	8924444.219	0.00	0.9983
1	CALC_CPI	AGE        39	21104.02156	1	8924444.204	0.00	0.9981
1	CALC_CPI	STD_YYYY   2002	-16791.17764	1	1029.07981	-16.32	0.0000
1	CALC_CPI	STD_YYYY   2003	-14332.50773	1	1015.028264	-14.12	0.0000
1	CALC_CPI	STD_YYYY   2004	-12365.53006	1	1009.390721	-12.25	0.0000
1	CALC_CPI	STD_YYYY   2005	-12066.34541	1	1006.206658	-11.99	0.0000
1	CALC_CPI	STD_YYYY   2006	-10180.47215	1	1004.022003	-10.14	0.0000
1	CALC_CPI	STD_YYYY   2007	-7857.90199	1	1002.274509	-7.84	0.0000
1	CALC_CPI	STD_YYYY   2008	-6173.65299	1	1000.796782	-6.17	0.0000
1	CALC_CPI	STD_YYYY   2009	-6142.29846	1	999.4356916	-6.15	0.0000
1	CALC_CPI	STD_YYYY   2010	-5945.25020	1	998.1000149	-5.96	0.0000
1	CALC_CPI	STD_YYYY   2011	-5474.52696	1	997.101357	-5.49	0.0000
1	CALC_CPI	STD_YYYY   2012	-4632.71579	1	996.0329327	-4.65	0.0000
1	CALC_CPI	STD_YYYY   2013	-4340.17720	1	995.1040046	-4.36	0.0000
1	CALC_CPI	STD_YYYY   2014	-3545.84023	1	994.2626587	-3.57	0.0004
1	CALC_CPI	STD_YYYY   2015	-2930.66837	1	990.6913263	-2.96	0.0031
1	CALC_CPI	STD_YYYY   2016	-2804.04633	1	991.2891071	-2.83	0.0047
1	CALC_CPI	STD_YYYY   2017	-2246.37908	1	997.3882092	-2.25	0.0243
1	CALC_CPI	STD_YYYY   2018	-1946.53825	1	1022.994013	-1.90	0.0571
1	CALC_CPI	STD_YYYY   2019	0.00000	1

METRO	Effect	EVENT_TIME	AGE	STD_YYYY	Parameter	DF	Estimate	StandardizedEst	StdErr	tValue	Probt
0	Intercept				Intercept	1	48732.76449	0	823.2806409	59.19	0.0000
0	EVENT_TIME	-3			EVENT_TIME -3	1	978.6121821	0.009879003	87.79498495	11.15	0.0000
0	EVENT_TIME	-2			EVENT_TIME -2	1	1180.696301	0.011916745	82.27548832	14.35	0.0000
0	EVENT_TIME	0			EVENT_TIME 0	1	-6342.501979	-0.063938752	81.54170722	-77.78	0.0000
0	EVENT_TIME	1			EVENT_TIME 1	1	-14450.89523	-0.142883041	84.74715836	-170.52	0.0000
0	EVENT_TIME	2			EVENT_TIME 2	1	-16030.37882	-0.160069536	87.51245683	-183.18	0.0000
0	EVENT_TIME	3			EVENT_TIME 3	1	-16749.013	-0.167109196	91.07833505	-183.90	0.0000
0	EVENT_TIME	4			EVENT_TIME 4	1	-17470.41779	-0.17395332	95.08918929	-183.73	0.0000
0	EVENT_TIME	5			EVENT_TIME 5	1	-17515.59946	-0.174604706	99.91191774	-175.31	0.0000
0	EVENT_TIME	-1			EVENT_TIME -1	0	0	0
0	AGE		22		AGE        22	1	-23951.48593	-0.065256495	609.8976565	-39.27	0.0000
0	AGE		23		AGE        23	1	-21995.39602	-0.092291837	571.3578958	-38.50	0.0000
0	AGE		24		AGE        24	1	-19659.34367	-0.10864938	558.0981989	-35.23	0.0000
0	AGE		25		AGE        25	1	-17930.41885	-0.121767376	551.3697335	-32.52	0.0000
0	AGE		26		AGE        26	1	-16556.64999	-0.129777789	547.285133	-30.25	0.0000
0	AGE		27		AGE        27	1	-15569.85237	-0.136392915	544.2738391	-28.61	0.0000
0	AGE		28		AGE        28	1	-14651.6871	-0.137586654	541.9167938	-27.04	0.0000
0	AGE		29		AGE        29	1	-13539.7169	-0.132324866	539.8663674	-25.08	0.0000
0	AGE		30		AGE        30	1	-12261.00612	-0.122064747	537.9408	-22.79	0.0000
0	AGE		31		AGE        31	1	-10811.67684	-0.104877075	536.4959055	-20.15	0.0000
0	AGE		32		AGE        32	1	-9331.853508	-0.086131356	534.890358	-17.45	0.0000
0	AGE		33		AGE        33	1	-8047.966748	-0.069081574	533.0982466	-15.10	0.0000
0	AGE		34		AGE        34	1	-6737.015473	-0.051968364	531.201936	-12.68	0.0000
0	AGE		35		AGE        35	1	-5409.257986	-0.03622049	527.5083801	-10.25	0.0000
0	AGE		36		AGE        36	1	-4179.530482	-0.022558388	524.2580505	-7.97	0.0000
0	AGE		37		AGE        37	1	-3038.194123	-0.012260304	523.1339909	-5.81	0.0000
0	AGE		38		AGE        38	1	-1389.644267	-0.003768741	528.8018235	-2.63	0.0086
0	AGE		39		AGE        39	0	0	0
0	STD_YYYY			2002	STD_YYYY   2002	1	-16650.53095	-0.080354869	996.552424	-16.71	0.0000
0	STD_YYYY			2003	STD_YYYY   2003	1	-14286.00212	-0.092404093	987.432746	-14.47	0.0000
0	STD_YYYY			2004	STD_YYYY   2004	1	-12407.17001	-0.095531127	983.6120877	-12.61	0.0000
0	STD_YYYY			2005	STD_YYYY   2005	1	-11746.07587	-0.100170107	981.3609237	-11.97	0.0000
0	STD_YYYY			2006	STD_YYYY   2006	1	-10098.34299	-0.091994308	979.7390713	-10.31	0.0000
0	STD_YYYY			2007	STD_YYYY   2007	1	-8150.321888	-0.077817622	978.4025271	-8.33	0.0000
0	STD_YYYY			2008	STD_YYYY   2008	1	-6771.204	-0.06640663	977.2427986	-6.93	0.0000
0	STD_YYYY			2009	STD_YYYY   2009	1	-6417.486667	-0.063835716	976.1549733	-6.57	0.0000
0	STD_YYYY			2010	STD_YYYY   2010	1	-6649.554649	-0.066117202	975.0897738	-6.82	0.0000
0	STD_YYYY			2011	STD_YYYY   2011	1	-6103.361286	-0.054600383	974.3559472	-6.26	0.0000
0	STD_YYYY			2012	STD_YYYY   2012	1	-5366.468102	-0.042711614	973.5318086	-5.51	0.0000
0	STD_YYYY			2013	STD_YYYY   2013	1	-4708.935048	-0.032176676	972.8157802	-4.84	0.0000
0	STD_YYYY			2014	STD_YYYY   2014	1	-3626.385435	-0.020955799	972.0437018	-3.73	0.0002
0	STD_YYYY			2015	STD_YYYY   2015	1	-3132.691922	-0.014746356	969.0205189	-3.23	0.0012
0	STD_YYYY			2016	STD_YYYY   2016	1	-2790.127585	-0.00996555	969.1815037	-2.88	0.0040
0	STD_YYYY			2017	STD_YYYY   2017	1	-2300.810792	-0.00577359	974.814409	-2.36	0.0183
0	STD_YYYY			2018	STD_YYYY   2018	1	-1647.516408	-0.002507904	999.3242518	-1.65	0.0992
0	STD_YYYY			2019	STD_YYYY   2019	0	0	0
1	Intercept				Intercept	1	63280.45558	0	850.487093	74.40	0.0000
1	EVENT_TIME	-3			EVENT_TIME -3	1	994.3418626	0.008055379	101.1486802	9.83	0.0000
1	EVENT_TIME	-2			EVENT_TIME -2	1	1215.894272	0.009848127	95.45161366	12.74	0.0000
1	EVENT_TIME	0			EVENT_TIME 0	1	-6843.483011	-0.055291696	94.85320156	-72.15	0.0000
1	EVENT_TIME	1			EVENT_TIME 1	1	-17267.81526	-0.135461874	98.94973511	-174.51	0.0000
1	EVENT_TIME	2			EVENT_TIME 2	1	-19545.15122	-0.155900401	101.849245	-191.90	0.0000
1	EVENT_TIME	3			EVENT_TIME 3	1	-20748.45919	-0.165749227	106.0720896	-195.61	0.0000
1	EVENT_TIME	4			EVENT_TIME 4	1	-22258.63616	-0.177409728	110.9774559	-200.57	0.0000
1	EVENT_TIME	5			EVENT_TIME 5	1	-23210.70219	-0.185285274	116.9501046	-198.47	0.0000
1	EVENT_TIME	-1			EVENT_TIME -1	0	0	0
1	AGE		22		AGE        22	1	-37718.49845	-0.068051731	668.144711	-56.45	0.0000
1	AGE		23		AGE        23	1	-35748.9006	-0.101599922	599.7267032	-59.61	0.0000
1	AGE		24		AGE        24	1	-32940.93284	-0.126637202	576.2575101	-57.16	0.0000
1	AGE		25		AGE        25	1	-30165.94661	-0.147222185	564.6762622	-53.42	0.0000
1	AGE		26		AGE        26	1	-27665.34236	-0.16156506	557.9669949	-49.58	0.0000
1	AGE		27		AGE        27	1	-25559.05246	-0.171838212	553.2894143	-46.19	0.0000
1	AGE		28		AGE        28	1	-23612.52525	-0.173613597	549.8253636	-42.95	0.0000
1	AGE		29		AGE        29	1	-21537.26904	-0.166900789	546.9335527	-39.38	0.0000
1	AGE		30		AGE        30	1	-19520.55581	-0.155215734	544.2995877	-35.86	0.0000
1	AGE		31		AGE        31	1	-17306.22797	-0.135745804	542.2007501	-31.92	0.0000
1	AGE		32		AGE        32	1	-15069.50123	-0.113959816	539.9454294	-27.91	0.0000
1	AGE		33		AGE        33	1	-12922.59306	-0.092519626	537.4588657	-24.04	0.0000
1	AGE		34		AGE        34	1	-10691.52796	-0.07045734	534.77624	-19.99	0.0000
1	AGE		35		AGE        35	1	-8369.566899	-0.048821541	530.1984657	-15.79	0.0000
1	AGE		36		AGE        36	1	-6214.63083	-0.029862688	526.2921207	-11.81	0.0000
1	AGE		37		AGE        37	1	-4101.209234	-0.015029863	524.639301	-7.82	0.0000
1	AGE		38		AGE        38	1	-2134.609388	-0.005332029	529.9816263	-4.03	0.0001
1	AGE		39		AGE        39	0	0	0
1	STD_YYYY			2002	STD_YYYY   2002	1	-16791.17786	-0.058028694	1029.079642	-16.32	0.0000
1	STD_YYYY			2003	STD_YYYY   2003	1	-14332.50796	-0.068346207	1015.028098	-14.12	0.0000
1	STD_YYYY			2004	STD_YYYY   2004	1	-12365.53029	-0.072008031	1009.390556	-12.25	0.0000
1	STD_YYYY			2005	STD_YYYY   2005	1	-12066.34564	-0.07940633	1006.206493	-11.99	0.0000
1	STD_YYYY			2006	STD_YYYY   2006	1	-10180.47238	-0.072531178	1004.021839	-10.14	0.0000
1	STD_YYYY			2007	STD_YYYY   2007	1	-7857.902214	-0.059275162	1002.274346	-7.84	0.0000
1	STD_YYYY			2008	STD_YYYY   2008	1	-6173.653216	-0.048113405	1000.796619	-6.17	0.0000
1	STD_YYYY			2009	STD_YYYY   2009	1	-6142.298686	-0.048723706	999.4355283	-6.15	0.0000
1	STD_YYYY			2010	STD_YYYY   2010	1	-5945.250427	-0.047334151	998.0998518	-5.96	0.0000
1	STD_YYYY			2011	STD_YYYY   2011	1	-5474.527188	-0.040184299	997.1011941	-5.49	0.0000
1	STD_YYYY			2012	STD_YYYY   2012	1	-4632.716016	-0.030845737	996.0327699	-4.65	0.0000
1	STD_YYYY			2013	STD_YYYY   2013	1	-4340.177422	-0.025316809	995.103842	-4.36	0.0000
1	STD_YYYY			2014	STD_YYYY   2014	1	-3545.840451	-0.017710699	994.2624963	-3.57	0.0004
1	STD_YYYY			2015	STD_YYYY   2015	1	-2930.668589	-0.012085427	990.6911644	-2.96	0.0031
1	STD_YYYY			2016	STD_YYYY   2016	1	-2804.046552	-0.008861841	991.2889451	-2.83	0.0047
1	STD_YYYY			2017	STD_YYYY   2017	1	-2246.379304	-0.005073097	997.3880462	-2.25	0.0243
1	STD_YYYY			2018	STD_YYYY   2018	1	-1946.538475	-0.002698223	1022.993846	-1.90	0.0571
1	STD_YYYY			2019	STD_YYYY   2019	0	0	0

Now, there's a problem in glm results.

I have no idea why this problem happened.

Can you give any information or suggestion? Or is there any pages that I can see each steps of how weights are applied in the regression?

[jhmoon]

And I use SAS Enterprise Guide!

[Rick_SAS]

What version of SAS are you running? For example, submit the following statement and copy/paste the answer from the log:
%put &=SYSVLONG;

[jhmoon]

It's SAS Enterprise Guide 7.15.

[Rick_SAS]

We need the version of SAS, not the version of EG. Please submit the SAS statement in my previous post and report the results in the log.

[Rick_SAS]

To help us figure out the cause, can you please

1. Use a QUIT statement in the GLM step.

2. Modify the GLMSELECT step to explicitly specify the reference levels and the GLM parameterization. 

CLASS EVENT_TIME(REF='-1') AGE(REF='39') STD_YYYY(REF='2019') / param=GLM;

Do you get the same parameter estimates?

[Rick_SAS]

This could be a data-dependent problem caused by having the same (or very small) weights for the STD_YYYY variable. We should look at the distribution of weights for each level of the STD_YYYY variable. Please run the following and post the results.

proc means DATA=IPW.TREAT_BASE_SAMPLE_CHT70 ;
   where not cmiss(Event_time, Age);
   class STD_YYYY;
   var C1_WEIGHT;
run;

proc sgplot DATA=IPW.TREAT_BASE_SAMPLE_CHT70 ;
   where not cmiss(Event_time, Age);
   vbox C1_WEIGHT / category= STD_YYYY;
run;

[jhmoon]

Sorry for late replying.
I have to export the result file and it takes several days.
I'll reply right after the file is exported.

[jhmoon]

Thank you for replying and suggestion, Rick.

Here's the result for your suggestions.

1. SAS version

%put &=SYSVLONG;

--> SYSVLONG=9.04.01M8P011823

2.  To help us figure out the cause, can you please

1. Use a QUIT statement in the GLM step.

2. Modify the GLMSELECT step to explicitly specify the reference levels and the GLM parameterization. 

CLASS EVENT_TIME(REF='-1') AGE(REF='39') STD_YYYY(REF='2019') / param=GLM;

Do you get the same parameter estimates?

--> yes, it shows exactly same results.

3.  "This could be a data-dependent problem caused by having the same (or very small) weights for the STD_YYYY variable. We should look at the distribution of weights for each level of the STD_YYYY variable. Please run the following and post the results."

proc means DATA=IPW.TREAT_BASE_SAMPLE_CHT70 ;
   where not cmiss(Event_time, Age);
   class STD_YYYY;
   var C1_WEIGHT;
run;

proc sgplot DATA=IPW.TREAT_BASE_SAMPLE_CHT70 ;
   where not cmiss(Event_time, Age);
   vbox C1_WEIGHT / category= STD_YYYY;
run;

--> Here's my code and results below:

/*RICK'S SOLUTION TO CHECK*/
PROC MEANS DATA=IPW.TREAT_BASE_SAMPLE_CHT70 ;
	WHERE NOT CMISS(EVENT_TIME, AGE);
	CLASS STD_YYYY;
	VAR C1_WEIGHT;
RUN;

ODS GRAPHICS ON / OBSMAX=5882292;

PROC SGPLOT DATA=IPW.TREAT_BASE_SAMPLE_CHT70 ;
	WHERE NOT CMISS(EVENT_TIME, AGE);
	VBOX C1_WEIGHT / CATEGORY= STD_YYYY;
RUN;

In addition, I follow your suggestion with AGE instead of STD_YYYY

 

You mentioned in last replying that it could be caused by having same weight.

But, I am still wondering why there are different results between 'glm' and 'glmselect'.

Even if it is a problem of weights, I think both codes should report the same results.

Thank you

[Rick_SAS]

Thanks for the additional information.

• It looks like you are using 9.4M8, which is the latest software.
• For several categories, the weights are constant, but the constant is not zero.

I do not know why you are seeing this issue. I suggest you contact SAS Technical Support and work with them to create an example that the developers at SAS can use to reproduce the problem.

> Even if it is a problem of weights, I think both codes should report the same results.

Yes, you are correct. The two procedure calls should produce the same results, as far as I can tell.

[jhmoon]

Thank you for your help, RIck.
Then, I will contact Technical Support team, following your suggestion.