Error with weight statement in proc lifetest

[marghe]

Hello all, 

I want to use IPW weights in my survival analysis to get the Kaplan-Meier curve (add the weight statement)  but when I run the procedure I get 

ERRROR: Domain error

ERROR: Floating point exception

Is there anyone that can fix it? I need it for my thesis!!

Thanks,

Margherita

[Rick_SAS]

First, check the quality of your data. Are the weights all positive? To get basic statistics for the weights, run

proc means data=MyData;

var MyWeightVar;

run;

If your weights are valid, then post some sample data and the SAS code that you are using so that we can see what you are doing.

[marghe]

I have two sets of weights, normal weights and stabilized weights, here is all the information about my data:

NORMAL WEIGHTS

Variabile di analisi: iptw

N            Average        Std dev         Min              Max

1400

1.8026681

0.9106701

1.0157942

6.1039446

STABILIZED WEIGHTS

Variabile di analisi: iptws

N             Average         Std dev         Min             Max

1400

0.9536681

0.3630199

0.4093205

2.5291374

As you can see they are all positive and not so small, so I don't know where the problem is 😞

Here is my dataset ordered by iptws (stabilized weights)

count         trt              sex         age         V1         V2       V3          V4         V5       event_time status   _LEVEL_   logit_ps ps iptw iptws

390	Treatment	M	74	7.25	25.9	3	20	2880	2	1	Treatment	0.83789	0.69802	1.43262	0.40932
344	Treatment	M	60	7.35	28.41	2	20	2958	3	1	Treatment	0.83777	0.69799	1.43268	0.40934
374	Treatment	M	61	7.03	28.32	3	20	2936	3	1	Treatment	0.81985	0.69421	1.44050	0.41157
179	Treatment	M	76	8.01	26.44	2	20	2741	3	1	Treatment	0.79273	0.68842	1.45261	0.41503
68	Treatment	M	64	8.87	27.89	2	20	2766	4	1	Treatment	0.77461	0.68452	1.46088	0.41740
9	Treatment	M	70	8.57	26.89	2	20	2529	3	1	Treatment	0.75219	0.67966	1.47133	0.42038
159	Treatment	M	77	8.22	27.41	3	20	2760	1	1	Treatment	0.73466	0.67583	1.47967	0.42276
12	Treatment	M	52	7.65	25.75	2	19	2893	2	1	Treatment	0.70801	0.66996	1.49262	0.42646
293	Treatment	M	73	8.37	29.79	3	20	2959	2	1	Treatment	0.70024	0.66824	1.49647	0.42756
169	Treatment	F	63	7.38	25.56	3	20	1707	3	1	Treatment	0.67746	0.66317	1.50791	0.43083
297	Treatment	M	79	8.74	29.3	2	20	2706	4	0	Treatment	0.66387	0.66013	1.51486	0.43282
166	Treatment	M	66	7.27	25.24	3	19	2830	2	1	Treatment	0.64981	0.65697	1.52215	0.43490
190	Treatment	M	66	8.83	29.26	3	20	2443	2	1	Treatment	0.64719	0.65638	1.52351	0.43529
354	Treatment	M	55	7.93	26.08	3	19	2798	2	1	Treatment	0.64312	0.65546	1.52565	0.43590
44	Treatment	F	52	7.54	28.17	3	20	1778	3	1	Treatment	0.63979	0.65471	1.52741	0.43640
278	Treatment	M	77	7.76	25.36	3	19	2958	2	1	Treatment	0.61542	0.64918	1.54041	0.44012
184	Treatment	M	76	8.2	25.44	3	19	2990	1	0	Treatment	0.61228	0.64846	1.54211	0.44060
210	Treatment	M	52	7.99	28.47	3	19	2991	2	1	Treatment	0.60557	0.64693	1.54576	0.44165
283	Treatment	M	52	7.44	28.47	4	19	2990	2	1	Treatment	0.59696	0.64496	1.55048	0.44300
389	Treatment	F	56	8.85	28.2	3	20	1779	4	1	Treatment	0.59534	0.64459	1.55137	0.44325
360	Treatment	M	60	8.69	27.01	2	19	2779	1	1	Treatment	0.59388	0.64426	1.55218	0.44348
235	Treatment	F	62	7.38	27.62	3	20	1596	3	1	Treatment	0.59351	0.64417	1.55239	0.44354
124	Treatment	M	52	7.48	26.85	3	19	2497	3	1	Treatment	0.58766	0.64283	1.55562	0.44446
261	Treatment	F	64	8.43	27.8	3	20	1767	4	0	Treatment	0.58588	0.64242	1.55661	0.44475
356	Treatment	F	57	7.97	29.4	2	20	1741	2	1	Treatment	0.58467	0.64214	1.55729	0.44494

and the procedure I want to run

proc lifetest data=iptw_db atrisk plots=survival(failure cl);
          strata trt;
          weight iptw; *the variable is iptws for stabilized weights;
          time event_time*status(0);
run;

ERROR: Domain error.

ERROR: Floating point exception.

 Many thanks!!

[Rick_SAS]

You must be running an old version of SAS. Your program runs without error in SAS 940m5.  

I found a SAS NOTE that talks about an error when using the STRATA statement in PROC LIFEREG and also requesting the survival plot. Try removing the survival plot request (get rid of the PLOTS= option) and see if that enables your analysis to complete.

[marghe]

Thank you! I tried to get rid of the PLOTS=.. but I got the same error; in any case, unfortunately, I am interested in the weighted survival plot. My SAS version is 9.4 M5 on SAS Studio. I have to find a solution.. maybe writing che code without any procedure.

[Rick_SAS]

I'm confused by your statement that you are running SAS 9.4 M5 because that is what I am running.

To make sure that we are running the same data, can you run the following and report what happens?

data iptw_db;
input count trt $ sex $ age V1 V2 V3 V4 V5 event_time status logit_ps ps iptw iptws;
datalines;
390   Treatment   M   74  7.25  25.9  3   20  2880  2   1   0.83789 0.69802 1.43262 0.40932
344   Treatment   M   60  7.35  28.41 2   20  2958  3   1   0.83777 0.69799 1.43268 0.40934
374   Treatment   M   61  7.03  28.32 3   20  2936  3   1   0.81985 0.69421 1.44050 0.41157
179   Treatment   M   76  8.01  26.44 2   20  2741  3   1   0.79273 0.68842 1.45261 0.41503
68    Treatment   M   64  8.87  27.89 2   20  2766  4   1   0.77461 0.68452 1.46088 0.41740
9     Treatment   M   70  8.57  26.89 2   20  2529  3   1   0.75219 0.67966 1.47133 0.42038
159   Treatment   M   77  8.22  27.41 3   20  2760  1   1   0.73466 0.67583 1.47967 0.42276
12    Treatment   M   52  7.65  25.75 2   19  2893  2   1   0.70801 0.66996 1.49262 0.42646
293   Treatment   M   73  8.37  29.79 3   20  2959  2   1   0.70024 0.66824 1.49647 0.42756
169   Treatment   F   63  7.38  25.56 3   20  1707  3   1   0.67746 0.66317 1.50791 0.43083
297   Treatment   M   79  8.74  29.3  2   20  2706  4   0   0.66387 0.66013 1.51486 0.43282
166   Treatment   M   66  7.27  25.24 3   19  2830  2   1   0.64981 0.65697 1.52215 0.43490
190   Treatment   M   66  8.83  29.26 3   20  2443  2   1   0.64719 0.65638 1.52351 0.43529
354   Treatment   M   55  7.93  26.08 3   19  2798  2   1   0.64312 0.65546 1.52565 0.43590
44    Treatment   F   52  7.54  28.17 3   20  1778  3   1   0.63979 0.65471 1.52741 0.43640
278   Treatment   M   77  7.76  25.36 3   19  2958  2   1   0.61542 0.64918 1.54041 0.44012
184   Treatment   M   76  8.2   25.44 3   19  2990  1   0   0.61228 0.64846 1.54211 0.44060
210   Treatment   M   52  7.99  28.47 3   19  2991  2   1   0.60557 0.64693 1.54576 0.44165
283   Treatment   M   52  7.44  28.47 4   19  2990  2   1   0.59696 0.64496 1.55048 0.44300
389   Treatment   F   56  8.85  28.2  3   20  1779  4   1   0.59534 0.64459 1.55137 0.44325
360   Treatment   M   60  8.69  27.01 2   19  2779  1   1   0.59388 0.64426 1.55218 0.44348
235   Treatment   F   62  7.38  27.62 3   20  1596  3   1   0.59351 0.64417 1.55239 0.44354
124   Treatment   M   52  7.48  26.85 3   19  2497  3   1   0.58766 0.64283 1.55562 0.44446
261   Treatment   F   64  8.43  27.8  3   20  1767  4   0   0.58588 0.64242 1.55661 0.44475
356   Treatment   F   57  7.97  29.4  2   20  1741  2   1   0.58467 0.64214 1.55729 0.44494
;

proc lifetest data=iptw_db atrisk plots=survival(failure cl);
          strata trt;
          weight iptw; *the variable is iptws for stabilized weights;
          time event_time*status(0);
run;

[marghe]

This code produces the plot I need but the dataset I copied in the previous post is just a portion of the whole dataset. So, the error must be related to the dataset.

[ashishpople1562]

I am facing the similar issue, Domain error , floating point exception. Can you tell what was the problem with the dataset that you were unable to run analysis on entire data. It would be great if you reply ASAP. 

Thanks in advance

[Rick_SAS]

Do you get the error for the data set in this thread?  

If you get the error for some data but not for other data, then something about the data is causing the problem. Sometimes you can determine the problem by running the program on a subset of the data. For example, if the problem goes away when you use 
proc lifetest data=Have(OBS=500);

then that is valuable information.  

If you can post your data here, then please do. If you cannot post your data because it is too large (or proprietary), then open a ticket with SAS Technical Support. They can make arrangements to obtain the data.

[ashishpople1562]

I cannot share the data due to certain constraints, actual data has more than 50k observations, we tried running with (obs=500) and (obs=1000) the model gave output. We also tried running the model with weight = 1 still we faced the domain error.