[Rd] Floating point issue

Martin Maechler m@ech|er @end|ng |rom @t@t@m@th@ethz@ch
Tue Jul 12 11:40:25 CEST 2022


>>>>> Martin Maechler 
>>>>>     on Mon, 11 Jul 2022 10:24:58 +0200 writes:

[............]

    > 10 days ago I also had a colleague at ETH, involved in teaching stats,
    > wanting to submit a bug report about a model fit in nlme which
    > gave different results (well, a confidence interval of a
    > correlation which was not at all significant (and estimated
    > close to zero) changed from +/-0.45 to +/-0.85 ...

(that was all the case, indeed)

    > ... and when I asked for sessionInfo() etc,
    > the result was that on the  M1  the result differed from all
    > other platforms.

when now rechecking and looking at the example, I've found to my
embarrassment, that the above paragraph (which *was* our conclusion)
has really not been the case.
Rather, as I've now reconfirmed, it seems that *only* on Windows
(from the platforms we looked at), the result differed from the others.

I.e., my nlme - example has been completely irrelevant to the
discussion here.
I do apologize!

Martin



More information about the R-devel mailing list