Re: [lmi] Anomalies indicated by 'rate_table_tool'

[Greg Chicares]

On 2019-03-07 22:47, Vadim Zeitlin wrote:
> On Thu, 7 Mar 2019 21:35:33 +0000 Greg Chicares <address@hidden> wrote:
> 
> GC> Vadim--I'm not sure which is anomalous, the table or the data, but
> GC> please try this test case...
[...]
>  I didn't look at the tables yet, but the logical conclusion is that it's
> the tool itself which must be buggy, right? If so, I guess I should debug
> these errors and I'll do it a.s.a.p., although probably not today and, if
> you could confirm that this is not too urgent, maybe even on Monday.

Sorry, my mistake--false alarm. Everything seems to be okay.

A (binary) database written by 'rate_table_tool' should pass its '--verify'
tests. But this database is dated 2003-09-22, so it certainly was not written
by 'rate_table_tool'. It must have been written by code that was published by
the Society of Actuaries in the 1990s. We already know that old code was not
really dependable; here, 'rate_table_tool' is simply diagnosing some of its
faults, as we have seen in the past, e.g., here:

  https://lists.nongnu.org/archive/html/lmi/2017-01/msg00001.html
| [...] whether tables downloaded from the SOA validate. [...]
[ [...] this work would have been done in the 1990s, and
| presumably they all used the same early version of the SOA program.
|
| - 'qx_ann' is less perfect, e.g.:
|       Verification failed for table #819: After loading and saving
|         the original table binary contents differed.
|       Table #910 specifies 5 decimals, but 6 were necessary

I can produce a 'sample.{dat,ndx}' binary database that does pass '--verify',
by using 'rate_table_tool' to extract all files and then to merge them all
into a new database.