1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
|
What are the differences between this testall.save and the one in
survival3 on statlib---
A "diff" gives a file with 850 lines! But the majority of the differences
are due to formatting differences between two version of S.
Line 110: Residuals for fit0. These have changed. Before, the fit was
with the Efron approx but residuals computed using Breslow. Now both use
Efron.
Line 243: Same issue, with the multiple event model.
263: ""
Line 447-537: A new test has been inserted.
In the parametric models, there are lots of changes due to formatting. For
the conconcted Fleming example, I get slightly different coefs, due to a
change in initial values logic.
---------
Further notes 5/16/94
Changed to 2 significant digits in the p value; leads to lots of changes
Added the Wald test to summary.coxph printout.
The "fleming" data is now referred to as "ovarian" data.
The expected survival from the "jasa1" test differs from before, for 2
reasons. The routine now does per year interpolation by default, but
more importantly I am now more careful about the US rate tables, ie. a
person moves to a new column of rates on their birthday rather than on
Jan 1 of the relevant year. The more nitpicky test of the new code in
doexpect.s makes me fairly certain of the new calculations.
-----
I have updated my testall script, and compared answers on 12/95.
New tests: infcox.s test the warning message for infinity, when 2
vars are infinite
doovarian.s added a test that robust=T and cluster() give the
same answer
docancer.s added a test of subscripting survival curves
expect3.s tests of survexp, with a Cox model
singtest.s test the 'singular=ok' option
Changes: The "beta may be infinite" message now has a semi-colon instead
of a comma
Statsci changed the # of digits on all the survreg output.
Added the ICC to multiple-event Cox models.
More "range" to "all.equal" changes.
--------
Notes on 6/96
Fixed the data error in the bladder data, where the 'size' and 'number'
variable names had been switched. This leads to lots of ignorable
differences. Also changed 'residual.dz' to 'resid.ds' to match my
course notes.
Removed the ICC from cox models until further testing is possible.
The survdiff program now has more complete output.
-------------
Notes on 9/98
A change in the convergence criteria for coxph changes lots of results
trivially.
Some new tests for survreg and expected survival.
|
