[R] Slightly OT - use of R

[Martin Maechler]

>>>>> "BDR" == Prof Brian Ripley <ripley at stats.ox.ac.uk>
>>>>>     on Mon, 30 Jul 2007 11:13:47 +0100 (BST) writes:

    BDR> On Mon, 30 Jul 2007, ted.harding at nessie.mcc.ac.uk wrote:
    >> On 30-Jul-07 08:28:15, John Logsdon wrote:
    >>> I am trying to get a measure of how R compares in usage as a
    >>> statistical platform compared to other software. I would guess
    >>> it is the most widely used among statisticians at least by
    >>> virtue of it being open source.

    BDR> I don't think that is the main reason.  Most of the R users I know 
    BDR> migrated from commercial statistical software for reasons other than cost.
    BDR> (Cross-platform availability has been one major reason.)

much of this is true here (Switzerland) as well.
{And some have *not* migrated because R is Free Software, but
 that's really another story}

Note however that the (non-PhD-graduate) students we teach here
would not be urged to using R if it was not the combination of
its quality and its Free Software state.
And I have had several acquaintances who have only started using
R because they could get it so easily and quickly, and they have
changed to using R as their main computational/statistical
software tool.

    >>> But is there any study to which I can refer? By asking this
    >>> list I am not exactly adopting a rigorous approach!
    >> 
    >> I don't know about that -- my own expectation would be that
    >> serious users of R are likely to be subscribers to the list.
    >> 
    >> So maybe a good answer to your question would be the number
    >> of subscribers (which I'm sure Martin Maechler can find out).
    >> Of course, some people will have subscribed under more than
    >> one email address, so that would somewhat over-estimate the
    >> number of people who subscribe. But it can be traded off
    >> (to a somewhat unknown extent) against R users who do not
    >> subscribe.

    BDR> I think it would be a seriously biased estimate.
    BDR> Few of our hundreds of student users will be subscribed to R-help 
    BDR> (since their first port of call for help is local).
    BDR> Also, we get quite a lot of postings via the gmane and nabble gateways.

Yes, yes, yes.
The exact same situation here and I'd believe in many places.

And the problem with the bias ('factor' rather than 'offset' I'd say)
is that it has been changing over time - I'd guess increasing pretty
dramatically.

My very wild subjective guess would be that 

   #{statisticians seriously using R} / 
   #{R-help subscribers} =  
    =  N_t / n_t

is nowadays well over 20, maybe even over 100,
of course depending on the definition of the numerator N_t.

I could construct a very accurate time-series for n_t,
but since I agree with Brian,
I haven't done so for several years.

Note that  n_{t = 2007-07-30, 07:00} = 5559

    >> More to the point, though, is what you mean by "usage".
    >> If you simply mean "people who use", that's a matter of
    >> counting (one way or another). But there's "use" and "use".

    BDR> Indeed.

"amen" - Martin