Re: [Emacs-diffs] master 2667d5c: Add new functions for the root mean square of a (Calc) vector

Re: [Emacs-diffs] master 2667d5c: Add new functions for the root mean square of a (Calc) vector

[Stefan Monnier]

-*- mode: nit-pick -*-

> +Another commonly used mean, the RMS (root-mean-square), can be computed
> +for a vector of numbers by using the @kbd{u R}

I must admit that I don't use Calc very often, and read its manual even
less often, so clearly I'm taking this out of context, but in any case:

To me a "vector" has a rather specific meaning, and taking the root mean
square of a vector doesn't make much sense.  Instead, I like to take the
rms of a *set* of numbers (which may be provided by packaging them in
a vector, of course).


        Stefan

Re: [Emacs-diffs] master 2667d5c: Add new functions for the root mean square of a (Calc) vector

[David Kastrup]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:

> -*- mode: nit-pick -*-
>
>> +Another commonly used mean, the RMS (root-mean-square), can be computed
>> +for a vector of numbers by using the @kbd{u R}
>
> I must admit that I don't use Calc very often, and read its manual even
> less often, so clearly I'm taking this out of context, but in any case:
>
> To me a "vector" has a rather specific meaning, and taking the root mean
> square of a vector doesn't make much sense.  Instead, I like to take the
> rms of a *set* of numbers (which may be provided by packaging them in
> a vector, of course).

Why would you ever want to do that?  It may at best make sense for a
finite _multiset_ (which can conveniently be represented by vector
elements).  The only difference then is that the members of a multiset
have no order.  But once you move to infinite measures as the base, RMS
is still well-defined but you'd rarely want to remove the order by only
viewing cumulative density functions.

At any rate, you could do s/vector of numbers/the numbers in a vector/
but indeed that may suggest that it may not count 1 twice in [1;2;1]
because the "numbers" are just 1 and 2.  So one would probably rather do
s/vector of numbers/the vector elements/.

-- 
David Kastrup

Re: [Emacs-diffs] master 2667d5c: Add new functions for the root mean square of a (Calc) vector

[Stefan Monnier]

> Why would you ever want to do that?  It may at best make sense for a
> finite _multiset_ (which can conveniently be represented by vector

Right, sorry, I meant multiset.


        Stefan

Re: [Emacs-diffs] master 2667d5c: Add new functions for the root mean square of a (Calc) vector

[Jay Belanger]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:

>> +Another commonly used mean, the RMS (root-mean-square), can be computed
>> +for a vector of numbers by using the @kbd{u R}

This documentation was already in the Calc manual, simply commented out.
I uncommented it and put in the (new) correct keybinding.

> To me a "vector" has a rather specific meaning, and taking the root mean
> square of a vector doesn't make much sense.  Instead, I like to take the
> rms of a *set* of numbers (which may be provided by packaging them in
> a vector, of course).

Yes; what's more, the documentation here doesn't match the documentation of the
other statistics functions.  I changed it to match the others, and now
it refers to the rms of data values rather than a vector.

Jay

Re: [Emacs-diffs] master 2667d5c: Add new functions for the root mean square of a (Calc) vector

[Eli Zaretskii]

> From: Stefan Monnier <monnier@IRO.UMontreal.CA>
> Date: Wed, 16 Sep 2015 09:21:40 -0400
> Cc: Jay Belanger <jay.p.belanger@gmail.com>
> 
> -*- mode: nit-pick -*-
> 
> > +Another commonly used mean, the RMS (root-mean-square), can be computed
> > +for a vector of numbers by using the @kbd{u R}
> 
> I must admit that I don't use Calc very often, and read its manual even
> less often, so clearly I'm taking this out of context, but in any case:
> 
> To me a "vector" has a rather specific meaning, and taking the root mean
> square of a vector doesn't make much sense.  Instead, I like to take the
> rms of a *set* of numbers (which may be provided by packaging them in
> a vector, of course).

As long as we are nit-picking, the result of root-mean-square
generally depends on the order of the values (due to finite precision
of FP calculations).  So it's not really an unordered set.

Re: [Emacs-diffs] master 2667d5c: Add new functions for the root mean square of a (Calc) vector

[David Kastrup]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Stefan Monnier <monnier@IRO.UMontreal.CA>
>> Date: Wed, 16 Sep 2015 09:21:40 -0400
>> Cc: Jay Belanger <jay.p.belanger@gmail.com>
>> 
>> -*- mode: nit-pick -*-
>> 
>> > +Another commonly used mean, the RMS (root-mean-square), can be computed
>> > +for a vector of numbers by using the @kbd{u R}
>> 
>> I must admit that I don't use Calc very often, and read its manual even
>> less often, so clearly I'm taking this out of context, but in any case:
>> 
>> To me a "vector" has a rather specific meaning, and taking the root mean
>> square of a vector doesn't make much sense.  Instead, I like to take the
>> rms of a *set* of numbers (which may be provided by packaging them in
>> a vector, of course).
>
> As long as we are nit-picking, the result of root-mean-square
> generally depends on the order of the values (due to finite precision
> of FP calculations).  So it's not really an unordered set.

Calc generally calculates exact expressions unless you tell it not to.
At any rate, the "squares" (for complex numbers, they should just be the
sum of squares of real and imaginary part rather than a complex square)
are actually all non-negative, so there is not a lot of cancellation
going on.  The precision of your result will be very slightly better if
you do the summation from smallest to large.

At any rate, given the amount of nits to be found by mailing list
members, it might be prudent to employ the services of a debugger.

-- 
David Kastrup