R devel - Jun 2008 - qf with infinite df

Dear all,

I found the following behaviour
> rf(5,Inf,Inf)
[1] 1 1 1 1 1

but
> qf(0.1,Inf,Inf)
[1] NaN
Warning messages:
1: In qf(0.1, Inf, Inf) : value out of range in 'lgamma'
2: In qf(p, df1, df2, lower.tail, log.p) : NaNs produced


incidentally,

> pf(1.00000000000001,Inf,Inf)
[1] 1
> pf(1.0000000000000001,Inf,Inf)
[1] 0.5

Is this the expected behaviour?

Thanks

Simone
> R.version
               _
platform       i386-pc-mingw32
arch           i386
os             mingw32
system         i386, mingw32
status         Patched
major          2
minor          7.0
year           2008
month          04
day            22
svn rev        45451
language       R
version.string R version 2.7.0 Patched (2008-04-22 r45451)

platform       x86_64-unknown-linux-gnu
arch           x86_64
os             linux-gnu
system         x86_64, linux-gnu
status         Patched
major          2
minor          6.1
year           2008
month          01
day            17
svn rev        44036
language       R
version.string R version 2.6.1 Patched (2008-01-17 r44036)

______________________________________________________

Simone Giannerini
Dipartimento di Scienze Statistiche "Paolo Fortunati"
Universita' di Bologna
Via delle belle arti 41 - 40126 Bologna, ITALY
Tel: +39 051 2098262 Fax: +39 051 232153
http://www2.stat.unibo.it/giannerini/
______________________________________________________

	[[alternative HTML version deleted]]

On Thu, 5 Jun 2008, Simone Giannerini wrote:
> Dear all,
>
> I found the following behaviour
>
>> rf(5,Inf,Inf)
> [1] 1 1 1 1 1
>
> but
>
>> qf(0.1,Inf,Inf)
> [1] NaN
> Warning messages:
> 1: In qf(0.1, Inf, Inf) : value out of range in 'lgamma'
> 2: In qf(p, df1, df2, lower.tail, log.p) : NaNs produced
Could do better here.
> incidentally,
>
>
>> pf(1.00000000000001,Inf,Inf)
> [1] 1
>> pf(1.0000000000000001,Inf,Inf)
> [1] 0.5
>
> Is this the expected behaviour?
I think so. 1.0000000000000001 is the same as 1 in computer arithmetic, 
and pf(1, m, n) = 0.5 for all large m, n.
>
> Thanks
>
> Simone
>
>> R.version
>               _
> platform       i386-pc-mingw32
> arch           i386
> os             mingw32
> system         i386, mingw32
> status         Patched
> major          2
> minor          7.0
> year           2008
> month          04
> day            22
> svn rev        45451
> language       R
> version.string R version 2.7.0 Patched (2008-04-22 r45451)
>
> platform       x86_64-unknown-linux-gnu
> arch           x86_64
> os             linux-gnu
> system         x86_64, linux-gnu
> status         Patched
> major          2
> minor          6.1
> year           2008
> month          01
> day            17
> svn rev        44036
> language       R
> version.string R version 2.6.1 Patched (2008-01-17 r44036)
>
> ______________________________________________________
>
> Simone Giannerini
> Dipartimento di Scienze Statistiche "Paolo Fortunati"
> Universita' di Bologna
> Via delle belle arti 41 - 40126 Bologna, ITALY
> Tel: +39 051 2098262 Fax: +39 051 232153
> http://www2.stat.unibo.it/giannerini/
> ______________________________________________________
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

Thank you, Simone,
>>>>> "SG" == Simone Giannerini <sgiannerini at
gmail.com>
>>>>>     on Thu, 5 Jun 2008 16:49:25 +0200 writes:
    SG> Dear all, I found the following behaviour

    >> rf(5,Inf,Inf)
    SG> [1] 1 1 1 1 1

    SG> but

    >> qf(0.1,Inf,Inf)
    SG> [1] NaN Warning messages: 1: In qf(0.1, Inf, Inf) :
    SG> value out of range in 'lgamma' 2: In qf(p, df1, df2,
    SG> lower.tail, log.p) : NaNs produced


    SG> incidentally,


    >> pf(1.00000000000001,Inf,Inf)
    SG> [1] 1
    >> pf(1.0000000000000001,Inf,Inf)
    SG> [1] 0.5

    SG> Is this the expected behaviour?
    	   ^^^^
What is   "this"?

rf() and pf() are.

qf() may still be "expected" if you think a bit how it
might be computed.

OTOH, we could clearly do better there.
Notably, qf()'s behavior is really not good enough, for the case
of large df1 and/or df2.

I'll have a look but won't promise much for now.

Martin Maechler, ETH Zurich


    SG> Thanks

    SG> Simone

    >> R.version
    SG>                _ platform i386-pc-mingw32 arch i386 os
    SG> mingw32 system i386, mingw32 status Patched major 2
    SG> minor 7.0 year 2008 month 04 day 22 svn rev 45451
    SG> language R version.string R version 2.7.0 Patched
    SG> (2008-04-22 r45451)

    SG> platform x86_64-unknown-linux-gnu arch x86_64 os
    SG> linux-gnu system x86_64, linux-gnu status Patched major
    SG> 2 minor 6.1 year 2008 month 01 day 17 svn rev 44036
    SG> language R version.string R version 2.6.1 Patched
    SG> (2008-01-17 r44036)

    SG> ______________________________________________________

    SG> Simone Giannerini Dipartimento di Scienze Statistiche
    SG> "Paolo Fortunati" Universita' di Bologna Via delle
belle
    SG> arti 41 - 40126 Bologna, ITALY Tel: +39 051 2098262 Fax:
    SG> +39 051 232153 http://www2.stat.unibo.it/giannerini/
    SG> ______________________________________________________