BBC iplayer
 

On Wed, 22 Aug 2007 01:39:54 -0700, "
wrote:

However, I would suggest to you that, just possibly, the record
companies looked at quantitive data such as sales, applied DRM to
different titles in different territories, and performed some kind of
analysis to judge the effect of applying the protection.

Some toerag record company enabled DRM on a CD that I bought a couple
of years ago preventing me playing it either on my PC or in the
multichanger in my car. It took an hour or so to eventually bypass it
and burn my own CD. But it still wasted a valuable hour of my time
and hence in protest I've stopped buying any CD's.

So DRM has in my view been a huge success, it's drastically reduced my
music buying habit and hit the record companies where it hurts. I
doubt I'll buy another CD this decade plus I'm damn sure i won't be
using either XP or Vista to view any UK television content.

--

BBC iplayer
 

On 23 Aug, 11:00, Roderick Stewart
wrote:
In article .com,

wrote:
Firstly, as explained, different titles in different countries carry
DRM.

So, for example, Dido's CD in the UK had DRM. The same CD in Australia
did not. The Beatles "Let it be - Naked" CD had DRM in the USA, in the
UK it did not. These are just two examples I'm aware of.

So if, for example, more copies of Dido's CD are sold per head of
population in one of these countries, how can a page of mathematics tell
us that this isn't simply because of their musical tastes? It often
happens that an artiste is more popular in one place than another.

One CD tells you nothing.

Multiply that by 10s of countries, 100s of releases, and you have more
than enough data for ANOVA to discover the significance, or otherwise,
of applying DRM.

That's the whole point of ANOVA. It lets you discover if variable X is
significant, even in the presence of variables Y and Z.

I don't know whether they did this or not. What I'm arguing is that
it's perfectly possible, and the statistics work. It's not magic.

It sounds impressive, and I'm sure it's possible they did this, and I'm
equally sure that the statistics "work", in the sense that the
calculations produce numerical results. However if the calculations that
led to those results are based on a supposition that is unfounded (e.g.
identical musical tastes, or identical efficacy of an artiste's
publicity campaign, in different countries), then of what value are
they? The phrase "garbage in, garbage out" comes to mind.

The whole point of ANOVA is that the assumptions are exactly the
opposite to what you state, i.e. that all variables may be just that:
variable. That's the whole point.

You could, of course, read up on it - and find out that the complexity
of performing ANOVA on this kind of data is almost prohibitive - but
as we don't have the data we don't know exactly what kind of result
we'd get - it could be a simple "don't know from this data".

Cheers,
David.

BBC iplayer
 

In article . com,
wrote:
The whole point of ANOVA is that the assumptions are exactly the
opposite to what you state, i.e. that all variables may be just that:
variable. That's the whole point.

Yes, but in this case the relevant variables are not independent. They
coincide exactly. The presence or absence of DRM coincides exactly with
the country in which the CD is sold. Thus any attempt to examine
differences between sales in the two countries and trying to separate
the effect from differences in musical preferences in the two countries
appears analogous to trying to sample two superimposed signals with the
same frequency and phase and using the same frequency to sample them.

You could, of course, read up on it - and find out that the complexity
of performing ANOVA on this kind of data is almost prohibitive - but
as we don't have the data we don't know exactly what kind of result
we'd get - it could be a simple "don't know from this data".

I think I can see straight away that we *cannot* know from this data.
The worrying thing is that some people are impressed by statistics if
the calculations look complicated enough, and may assume that this gives
them greater validity than a simple rational argument in Plain English.

Rod.

BBC iplayer
 

On 23 Aug, 20:21, Roderick Stewart
wrote:
In article . com,

wrote:
The whole point of ANOVA is that the assumptions are exactly the
opposite to what you state, i.e. that all variables may be just that:
variable. That's the whole point.

Yes, but in this case the relevant variables are not independent. They
coincide exactly. The presence or absence of DRM coincides exactly with
the country in which the CD is sold. Thus any attempt to examine
differences between sales in the two countries and trying to separate
the effect from differences in musical preferences in the two countries
appears analogous to trying to sample two superimposed signals with the
same frequency and phase and using the same frequency to sample them.

You could, of course, read up on it - and find out that the complexity
of performing ANOVA on this kind of data is almost prohibitive - but
as we don't have the data we don't know exactly what kind of result
we'd get - it could be a simple "don't know from this data".

I think I can see straight away that we *cannot* know from this data.
The worrying thing is that some people are impressed by statistics if
the calculations look complicated enough, and may assume that this gives
them greater validity than a simple rational argument in Plain English.

But you can't make a rational argument when you've fundamentally
misunderstood what's happening.


Let me try to explain it this way.

Imagine you take a very large sample. Imagine you perfectly randomise
the decision to include DRM or not for each title in each country
individually. If the sample is large enough, and if the randomisation
is good enough, then you can see what effect, on average, the
inclusion of DRM has.

What ANOVA can tell you is if the sample _was_ large enough, and the
randomisation _was_ good enough - or if, in fact, any effect you see
could or is just down to the other variables which you didn't
randomise.

That's not a good definition of ANOVA, and not actually how it works,
but it's a way of thinking about its output in this case.

Really it's like a drugs trial - there could be many confounding and
related factors. The trick is randomisation and sample size.


You seam to be hung up on the idea of one title in two countries, one
with DRM and one without. Of course that, on its own, tells you very
little - just like a drugs trial with two patients wouldn't tell you
much either.


Cheers,
David.

BBC iplayer
 

In article .com,
wrote:
On 20 Aug, 22:30, Roderick Stewart
wrote:

It will only generate "extra" revenue if the sales you imagine have
been "forced" by its non-availability outnumber the sales prompted by
its availability. Would you care to outline a reliable objective
method for measuring these quantities?

Of course not, but through various audio/hacker forums, I've observed
the spread of DRM on audio CDs. It wasn't introduced on all titles in
all territories at the same time. It was introduced piecemeal. One
plausible explanation is that the record companies were testing the
water, and assessing the impact it had.

How do you employ DRM (as distinct from some daft form of anti-copy system)
with Red Book Audio CD? My understanding is that DRM implies you need a
system to 'enable' legal use, not just make cloning a pain for non-pro
pirates. I know there are various daft 'anti copy' systems for CD, making
them non Red Book. But I don't know about how you'd apply DRM.

I haven't bought a "pop" CD in years, but from what I read, DRM is now
widespread on such CDs. This suggests the impact the record companies
saw justified the price of applying it. I admit there are other
explanations, but this is a plausible one.

Given that the analogue output of a decent CD system have a dynamic range
well over 50dB greater than the dyanamics on most 'pop' CDs I can't see DRM
having the slightest value when someone can make an analogue-linked
copy.

"They are numbskulls and spent money on something that brought them zero
benefit in any way" is less plausible, though possible of course. A lot
of things happen because someone in some company has to be seen to be
doing something. Lots of people create/perpetuate jobs for themselves
where no actual useful work exists for them to do!

Indeed. No doubt many internal and consultancy reports get written, and
people remain employed, to keep the plates spinning. Nice to know that the
money we pay over the counter is so well used. :-)

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.audiomisc.co.uk/index.html
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html

BBC iplayer
 

In article .com,
wrote:
On 22 Aug, 20:32, Roderick Stewart
wrote:


That's the whole point of ANOVA. It lets you discover if variable X is
significant, even in the presence of variables Y and Z.


I don't know whether they did this or not. What I'm arguing is that it's
perfectly possible, and the statistics work. It's not magic.

It can work. But is assumes various things. e,g. that there is no
correlated variable which means your assumption about what causes any
statistical 'detection' isn't what you've assumed. Also the obvious one,
that the result may be chance despite seeming otherwise. Easy to be
misled by statistics if you don't fully understand the situation being
tested. So saying the method "does work" isn't a guarantee that it
actually 'proves' a given conclusion *is* the correct one. It "works"
as a way of making estimates of probability, and of reliability - on
the basis of some assumptions turning out to be correct.

However I've never seen any report that they've done what you describe in a
meaningful manner for this topic. If the people who run the 'media'
companies *really* understood these matters then I guess nothing they
ever released would be a flop. ;-

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.audiomisc.co.uk/index.html
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html

BBC iplayer
 

In article .com,
wrote:
You seam to be hung up on the idea of one title in two countries, one
with DRM and one without. Of course that, on its own, tells you very
little - just like a drugs trial with two patients wouldn't tell you
much either.

Any larger scale statistical sampling is going to be a multiple of that
same situation, as long as all copies of a given title in any given
country either have, or don't have, DRM.

If the presence or absence of DRM were truly randomised amongst a
selection of the same people with the same national preferences in
music, then I can see how it would be possible to extract some data
about any effect DRM might have on sales, all other things being equal.

But all other things are not equal. Different countries most definitely
do have national preferences in music (or any kind of art) - you only
have to listen to what they produce themselves, or look at the fact
that any individual artiste often has different levels of popularity in
different countries, and the same recording will often have very
different sales. This happens anyway, regardless of any technical
gizmos, regardless of differences in technology even.

If we were to look at the relative popularities in different countries
of a globally famous TV programme (pick your own example), and someone
were to suggest that any part of this could be attributed to the fact
that some countries used NTSC broadcast signals and some used PAL,
you'd think they were crazy. What you would need would be parallel
transmissions of the same programme to the same communities in the same
country but using different transmission systems, and *then* you might
be able to say the differences had something to do with the technology.
What we actually have is a great many TV programmes that are popular
everywhere regardless of the technology, some more popular in some
places than others, but the technology always being the same for a
whole country. If the type of technology coincides with an area of
common national preference, how can we say to which cause any
differences can be attributed?

Rod.

BBC iplayer
 

On 24 Aug, 12:30, Roderick Stewart
wrote:
In article .com,

wrote:
You seam to be hung up on the idea of one title in two countries, one
with DRM and one without. Of course that, on its own, tells you very
little - just like a drugs trial with two patients wouldn't tell you
much either.

Any larger scale statistical sampling is going to be a multiple of that
same situation, as long as all copies of a given title in any given
country either have, or don't have, DRM.

The "large scale" is in terms of number of countries, and number of
titles. Each title+country combination is one data point.

If the presence or absence of DRM were truly randomised amongst a
selection of the same people with the same national preferences in
music, then I can see how it would be possible to extract some data
about any effect DRM might have on sales, all other things being equal.

But all other things are not equal. Different countries most definitely
do have national preferences in music (or any kind of art) - you only
have to listen to what they produce themselves, or look at the fact
that any individual artiste often has different levels of popularity in
different countries, and the same recording will often have very
different sales. This happens anyway, regardless of any technical
gizmos, regardless of differences in technology even.

If we were to look at the relative popularities in different countries
of a globally famous TV programme (pick your own example), and someone
were to suggest that any part of this could be attributed to the fact
that some countries used NTSC broadcast signals and some used PAL,
you'd think they were crazy. What you would need would be parallel
transmissions of the same programme to the same communities in the same
country but using different transmission systems, and *then* you might
be able to say the differences had something to do with the technology.
What we actually have is a great many TV programmes that are popular
everywhere regardless of the technology, some more popular in some
places than others, but the technology always being the same for a
whole country. If the type of technology coincides with an area of
common national preference, how can we say to which cause any
differences can be attributed?

But if one title can have DRM in country 1, and no DRM in country 2 -
while another title can have no DRM in country 1, and DRM in country 2
- then that's exactly the same has transmitting some programmes in PAL
and some programmes in NTSC in the same country - i.e. divorcing the
system (or protection) from the country.

A given title will be intrinsically more popular in one country than
another, but if you have lots of titles, and randomise the application
of DRM amongst them by country properly, you can measure the effect,
if any, given enough data.

Cheers,
David.

BBC iplayer
 

On 24 Aug, 12:20, Jim Lesurf wrote:
In article .com,
wrote:

On 22 Aug, 20:32, Roderick Stewart
wrote:
That's the whole point of ANOVA. It lets you discover if variable X is
significant, even in the presence of variables Y and Z.
I don't know whether they did this or not. What I'm arguing is that it's
perfectly possible, and the statistics work. It's not magic.

It can work. But is assumes various things. e,g. that there is no
correlated variable which means your assumption about what causes any
statistical 'detection' isn't what you've assumed. Also the obvious one,
that the result may be chance despite seeming otherwise. Easy to be
misled by statistics if you don't fully understand the situation being
tested. So saying the method "does work" isn't a guarantee that it
actually 'proves' a given conclusion *is* the correct one. It "works"
as a way of making estimates of probability, and of reliability - on
the basis of some assumptions turning out to be correct.

Well, exactly - that's the output - the probability of what you're
seeing being chance (or 1 minus that = the probability of what you're
seeing being a real effect). The output is never 0 (or 1).


However I've never seen any report that they've done what you describe in a
meaningful manner for this topic.

No, me neither. What I'm taking issue with is Rod's assertion that
this is simply impossible.

If the people who run the 'media'
companies *really* understood these matters then I guess nothing they
ever released would be a flop. ;-

Ah, but that's art. I don't understand that at all.

But I know what I like ;-)

Cheers,
David.

BBC iplayer
 

In article .com,
wrote:
Any larger scale statistical sampling is going to be a multiple of that
same situation, as long as all copies of a given title in any given
country either have, or don't have, DRM.

The "large scale" is in terms of number of countries, and number of
titles. Each title+country combination is one data point.

If each "data point" tells us nothing, lots of the same kind of data point
will tell us lots of nothing. How is that better?

[...]
But if one title can have DRM in country 1, and no DRM in country 2 -
while another title can have no DRM in country 1, and DRM in country 2
- then that's exactly the same has transmitting some programmes in PAL
and some programmes in NTSC in the same country - i.e. divorcing the
system (or protection) from the country.

Exactly. It's equivalent to transmitting *some* programmes in one system
and *some* programmes in another. How is it possible to say that the reason
more people watched a particular programme wasn't simply that because they
liked the programme. It's an unknown effect, and no matter how many times
you multiply the unknown, it's still unknown. What would be needed to
compare systems would be to transmit the *same* programme to the *same*
viewers using different systems.

A given title will be intrinsically more popular in one country than
another, but if you have lots of titles, and randomise the application
of DRM amongst them by country properly, you can measure the effect,
if any, given enough data.

I still think there's a fundamental flaw in the logic here. Comparing
viewers' or listeners' preferences using different programme material, when
the difference in programme material corresponds exactly with a difference
in technology, fails to isolate the two effects and so cannot tell us
anything about a possible preference for the different technologies. It
doesn't matter how much data of this sort we manage to accumulate if it's
fundamentally flawed. If one comparison cannot tell us anything meaningful
because of the kind of comparison it is, then multiplying it by a hundred,
or a million, won't tell us anything more.

Rod.