Frequently in ciwah people say 'but what about the users without
JavaScript?', so I decided to do an experiment. It suggests 35%
internet users do not have JavaScript turned on in their browsers. I'd
appreciate it if people here had a look at my page on the experiment,
and tell me if they see any methodological flaws, before I go and
publicise it more widely:

http://www.safalra.com/hypertext/html/nojavascript.html

--- Safalra (Stephen Morley) ---
http://www.safalra.com/hypertext

On 5 Mar 2004 09:49:01 -0800, Safalra wrote:
[color=blue]
> http://www.safalra.com/hypertext/html/nojavascript.html[/color]

Interesting figure, what is your usual
demographic at that page? That would be
the single biggest source of possible bias.

Oh, and you miss-spelled MurkoSoft ..errr
MircoSoft, ..err _MicroSoft_ in the last part
of the page. ;-)

--
Andrew Thompson
* http://www.PhySci.org/ Open-source software suite
* http://www.PhySci.org/codes/ Web & IT Help
* http://www.1point1C.org/ Science & Technology

"Safalra" <usenet@safalra.com> wrote in message
news:c555467b.0403050949.114db5b6@posting.google.c om...[color=blue]
> Frequently in ciwah people say 'but what about the users without
> JavaScript?', so I decided to do an experiment. It suggests 35%
> internet users do not have JavaScript turned on in their browsers.[/color]

No, it suggests that 35% of SETI@home users (who visited the Error Codes
page between Feb. X and Feb X+14) do not have JavaScript turned on it thier
browsers. There is nothing that suggests visitors of this particular site
represent all internet users as a whole. Your experiment is geared towards
a very specific audience, and I would say is not diverse enough to conclude
that 35% of "internet users" do not have JavaScript turned on.

[color=blue]
> I'd
> appreciate it if people here had a look at my page on the experiment,
> and tell me if they see any methodological flaws, before I go and
> publicise it more widely:[/color]


Considering that there are millions of people browsing the web, I think a
sampling of 859 people is hardly enough to make any real conclusions. In
addition, you should be making use of the <noscript> element in your
experiment, rather than the hack you created, which never closes the opening
comment (thus, I'm not sure how consistent the output would be accross a
number of different browsers).

For example:
<script language="javascript">
document.write('<a href=setierrorsjs.txt>a new location</a>.');
</script>
<noscript>
<a href=setierrorsnojs.html>a new location</a>.
</noscript>

I think your experiment is flawed, and the results questionable.


Regards,
Peter Foti

"Peter Foti" <peter@Idontwantnostinkingemailfromyou.com> wrote in message
news:104hi5jfc71k202@corp.supernews.com...[color=blue]
> "Safalra" <usenet@safalra.com> wrote in message
> news:c555467b.0403050949.114db5b6@posting.google.c om...[color=green]
> > Frequently in ciwah people say 'but what about the users without
> > JavaScript?', so I decided to do an experiment. It suggests 35%
> > internet users do not have JavaScript turned on in their browsers.[/color]
>
> No, it suggests that 35% of SETI@home users (who visited the Error Codes
> page between Feb. X and Feb X+14) do not have JavaScript turned on it[/color]
thier[color=blue]
> browsers. There is nothing that suggests visitors of this particular site
> represent all internet users as a whole. Your experiment is geared[/color]
towards[color=blue]
> a very specific audience, and I would say is not diverse enough to[/color]
conclude[color=blue]
> that 35% of "internet users" do not have JavaScript turned on.
>
>[color=green]
> > I'd
> > appreciate it if people here had a look at my page on the experiment,
> > and tell me if they see any methodological flaws, before I go and
> > publicise it more widely:[/color]
>
>
> Considering that there are millions of people browsing the web, I think a
> sampling of 859 people is hardly enough to make any real conclusions.[/color]

This may not be true. When valid sampling techniques are used, as I recall,
the number of subjects required for a valid yes-or-no poll of U.S.
residents, for the confidence interval to be 5 percentage points, is
something around 1,000, as long as no demographic breakdowns are required..

Harlan Messinger wrote:[color=blue]
> "Peter Foti" <peter@Idontwantnostinkingemailfromyou.com> wrote in
> message news:104hi5jfc71k202@corp.supernews.com...[/color]
[snip][color=blue][color=green]
>> Considering that there are millions of people browsing the web, I
>> think a sampling of 859 people is hardly enough to make any real
>> conclusions.[/color]
>
> This may not be true. When valid sampling techniques are used, as I
> recall, the number of subjects required for a valid yes-or-no poll of
> U.S. residents, for the confidence interval to be 5 percentage
> points, is something around 1,000, as long as no demographic
> breakdowns are required.[/color]

True. Have a look at:
http://www.surveysystem.com/sscalc.htm

Sampling 1000 people, in a properly controlled way, can be very good indeed.

But if it is inherently biased, it is pretty well useless. Certainly, don't
just ring people up!

--
Barry Pearson
http://www.Barry.Pearson.name/photography/
http://www.BirdsAndAnimals.info/
http://www.ChildSupportAnalysis.co.uk/

"Harlan Messinger" <h.messinger@comcast.net> wrote in message
news:c2aiim$1q090o$1@ID-114100.news.uni-berlin.de...[color=blue]
>
> "Peter Foti" <peter@Idontwantnostinkingemailfromyou.com> wrote in message
> news:104hi5jfc71k202@corp.supernews.com...[color=green]
> > "Safalra" <usenet@safalra.com> wrote in message
> > news:c555467b.0403050949.114db5b6@posting.google.c om...[color=darkred]
> > > Frequently in ciwah people say 'but what about the users without
> > > JavaScript?', so I decided to do an experiment. It suggests 35%
> > > internet users do not have JavaScript turned on in their browsers.[/color]
> >
> > No, it suggests that 35% of SETI@home users (who visited the Error Codes
> > page between Feb. X and Feb X+14) do not have JavaScript turned on it[/color]
> thier[color=green]
> > browsers. There is nothing that suggests visitors of this particular[/color][/color]
site[color=blue][color=green]
> > represent all internet users as a whole. Your experiment is geared[/color]
> towards[color=green]
> > a very specific audience, and I would say is not diverse enough to[/color]
> conclude[color=green]
> > that 35% of "internet users" do not have JavaScript turned on.
> >
> >[color=darkred]
> > > I'd
> > > appreciate it if people here had a look at my page on the experiment,
> > > and tell me if they see any methodological flaws, before I go and
> > > publicise it more widely:[/color]
> >
> >
> > Considering that there are millions of people browsing the web, I think[/color][/color]
a[color=blue][color=green]
> > sampling of 859 people is hardly enough to make any real conclusions.[/color]
>
> This may not be true. When valid sampling techniques are used, as I[/color]
recall,[color=blue]
> the number of subjects required for a valid yes-or-no poll of U.S.
> residents, for the confidence interval to be 5 percentage points, is
> something around 1,000, as long as no demographic breakdowns are[/color]
required..

I have no knowledge of this one way or the other, so I really can't respond
with anything other than my own opinion. In this case, I don't think valid
sampling techniques were used. 859 of a very specific demographic does not
seem like a valid sampling to me, especially since it was over the course of
2 weeks (that's only about 61 clicks per day, out of millions). You may be
correct regarding the 1,000 subjects, but I can't say one way or the other.

I think a better study may have been to try implementing this on a number of
different websites. But even then, you have the problem that only certain
people will click on your link. This would be a tough experiment to do
right.

Best,
Peter

It seems "Andrew Thompson" wrote in
comp.infosystems.www.authoring.html:[color=blue]
>On 5 Mar 2004 09:49:01 -0800, Safalra wrote:
>[color=green]
>> http://www.safalra.com/hypertext/html/nojavascript.html[/color]
>
>Oh, and you miss-spelled MurkoSoft ..errr
>MircoSoft, ..err _MicroSoft_ in the last part
>of the page. ;-)[/color]

I'm looking at the 17:33 GMT version of the page, which shows
"Microsoft", the correct spelling. (There is no interior capital.)

Your mistaken correction is dated 16 minutes later.

Either Morley has a time machine, or someone's computer time
settings are wrong, or you are mistaken.

--
Stan Brown, Oak Road Systems, Cortland County, New York, USA
http://OakRoadSystems.com/
HTML 4.01 spec: http://www.w3.org/TR/html401/
validator: http://validator.w3.org/
CSS 2 spec: http://www.w3.org/TR/REC-CSS2/
2.1 changes: http://www.w3.org/TR/CSS21/changes.html
validator: http://jigsaw.w3.org/css-validator/

Disclosure: I'm not a professional statistician, but I teach
introductory statistics at the local community college.


It seems "Harlan Messinger" wrote in
comp.infosystems.www.authoring.html:[color=blue]
>It seems "Safalra" wrote in comp.infosystems.www.authoring.html:[color=green]
>>http://www.safalra.com/hypertext/html/nojavascript.html[/color][/color]
[color=blue]
> When valid sampling techniques are used, as I recall,
>the number of subjects required for a valid yes-or-no poll of U.S.
>residents, for the confidence interval to be 5 percentage points, is
>something around 1,000, as long as no demographic breakdowns are required..[/color]

You have the germ of the right idea here, but the details are not
quite right. And there's one crucial assumption that I do not think
is met by Morley's data.

In general in statistics, the size of the sample matters but the
size of the population does not, as long as the sample is less than
about 10% of the population. For example, a sample of 1000 people is
as good for describing Cortland, New York (pop 24,000) as for
describing the population of the world (pop > 6,000,000,000) -- IF
IT'S A RANDOM SAMPLE.

And there's the rub: if the sample is not random then you really
can't draw any conclusion from it.

When sampling yes/no data like "does this user use Javascript?", if
you have a random sample of about 1,000 people then you can be 95%
confident that you know the opinion of the whole population within 3
(not 5) percentage points either way. This "95% confidence interval
with 3 points margin of error" is customary in political polling,
though any other combination could also be used. (For the
mathematically inclined: the margin of error is inversely
proportional to the square root of the sample size. To cut the
margin of error in half your sample must be four times as large.)

Morley has data at
<http://www.safalra.com/hypertext/html/nojavascript.html> with
sample size 859, of whom 307 (35.7%) have disabled Javascript. If it
were a random sample, we could then say with 95% confidence that
35.7±3.2% of the population has Javascript disabled.

But again, the sample is almost certainly not random. For one thing
it's self selected, always a concern. For another, the subject
matter (SETI) of the test page appeals to only a tiny minority of
people or even of Web surfers. There are actually two _unverified_
assumptions here: that Morley's sample is a true random sample even
of SETI enthusiasts, and that SETI enthusiasts have the same level
of Javascript usage as Web surfers in general. Both seem
questionable to me, the second extremely so. A failure of either is
fatal.

Another sample of Morley's, mentioned on his page at
<http://www.xs4all.nl/~sbpoley/webmatters/whatnojs.html>, found 25%
non-Javascript users. Again, that calls the 36% figure into
question. Even 25% is too high, I suspect. Morley's pages tend to
appeal to people who think carefully about Web design, and the
evidence of the c.i.w.a.* newsgroups suggests that such people tend
to turn off Javascript though most Web surfers aren't even aware it
exists.

Regretfully, I have to say that Morley's figures are interesting,
and his sample size is not a problem; but the lack of randomness
means that no conclusions can be drawn. I have no immediate
suggestion how to get a random sample to answer his question, but
the folks at sci.stat.edu might be able to help.

--
Stan Brown, Oak Road Systems, Cortland County, New York, USA
http://OakRoadSystems.com/
HTML 4.01 spec: http://www.w3.org/TR/html401/
validator: http://validator.w3.org/
CSS 2 spec: http://www.w3.org/TR/REC-CSS2/
2.1 changes: http://www.w3.org/TR/CSS21/changes.html
validator: http://jigsaw.w3.org/css-validator/

It seems "Safalra" wrote in comp.infosystems.www.authoring.html:[color=blue]
>tell me if they see any methodological flaws, before I go and
>publicise it more widely:
>
>http://www.safalra.com/hypertext/html/nojavascript.html[/color]

Interesting and worthwhile idea, but for the results to have any
validity the sample must be random, which it's not. (More details in
another message in this thread.)

I confess I don't have any immediate ideas how to come up with a
good random sample simply by observing behavior of visitors to a
particular site.

By the way, a numerical correction: 307/(552+307) = 36% not 35%.

--
Stan Brown, Oak Road Systems, Cortland County, New York, USA
http://OakRoadSystems.com/
HTML 4.01 spec: http://www.w3.org/TR/html401/
validator: http://validator.w3.org/
CSS 2 spec: http://www.w3.org/TR/REC-CSS2/
2.1 changes: http://www.w3.org/TR/CSS21/changes.html
validator: http://jigsaw.w3.org/css-validator/

"Peter Foti" <peter@Idontwantnostinkingemailfromyou.com> wrote in message
news:104hk6pk4p9j9c5@corp.supernews.com...[color=blue]
> "Harlan Messinger" <h.messinger@comcast.net> wrote in message
> news:c2aiim$1q090o$1@ID-114100.news.uni-berlin.de...[color=green]
> >
> > "Peter Foti" <peter@Idontwantnostinkingemailfromyou.com> wrote in[/color][/color]
message[color=blue][color=green]
> > news:104hi5jfc71k202@corp.supernews.com...[color=darkred]
> > > "Safalra" <usenet@safalra.com> wrote in message
> > > news:c555467b.0403050949.114db5b6@posting.google.c om...
> > > > Frequently in ciwah people say 'but what about the users without
> > > > JavaScript?', so I decided to do an experiment. It suggests 35%
> > > > internet users do not have JavaScript turned on in their browsers.
> > >
> > > No, it suggests that 35% of SETI@home users (who visited the Error[/color][/color][/color]
Codes[color=blue][color=green][color=darkred]
> > > page between Feb. X and Feb X+14) do not have JavaScript turned on it[/color]
> > thier[color=darkred]
> > > browsers. There is nothing that suggests visitors of this particular[/color][/color]
> site[color=green][color=darkred]
> > > represent all internet users as a whole. Your experiment is geared[/color]
> > towards[color=darkred]
> > > a very specific audience, and I would say is not diverse enough to[/color]
> > conclude[color=darkred]
> > > that 35% of "internet users" do not have JavaScript turned on.
> > >
> > >
> > > > I'd
> > > > appreciate it if people here had a look at my page on the[/color][/color][/color]
experiment,[color=blue][color=green][color=darkred]
> > > > and tell me if they see any methodological flaws, before I go and
> > > > publicise it more widely:
> > >
> > >
> > > Considering that there are millions of people browsing the web, I[/color][/color][/color]
think[color=blue]
> a[color=green][color=darkred]
> > > sampling of 859 people is hardly enough to make any real conclusions.[/color]
> >
> > This may not be true. When valid sampling techniques are used, as I[/color]
> recall,[color=green]
> > the number of subjects required for a valid yes-or-no poll of U.S.
> > residents, for the confidence interval to be 5 percentage points, is
> > something around 1,000, as long as no demographic breakdowns are[/color]
> required..
>
> I have no knowledge of this one way or the other, so I really can't[/color]
respond[color=blue]
> with anything other than my own opinion. In this case, I don't think[/color]
valid[color=blue]
> sampling techniques were used. 859 of a very specific demographic does[/color]
not[color=blue]
> seem like a valid sampling to me,[/color]

The latter is a problem if you are correct, but if the sampling were valid,
then 859 would probably be sufficient for a ballpark estimate.

In article <104hi5jfc71k202@corp.supernews.com>,
"Peter Foti" <peter@Idontwantnostinkingemailfromyou.com> writes:
[color=blue]
> represent all internet users as a whole. Your experiment is geared towards
> a very specific audience, and I would say is not diverse enough to conclude
> that 35% of "internet users" do not have JavaScript turned on.[/color]

Indeed.

But his sample is nevertheless probably better than the sometimes-cited
"thecounter.com", which has a sample bias driven not only by demographic
(as is Safalra's) but also by technology.
[color=blue]
> Considering that there are millions of people browsing the web, I think a
> sampling of 859 people is hardly enough to make any real conclusions. In[/color]

That sample size is not inherently problematic. Sample bias is the issue,
and a million visitors are neither better nor worse than 859 if the sample
is biased.

--
Nick Kew

Stan Brown <the_stan_brown@fastmail.fm> wrote in message news:<MPG.1ab2b633626f4bad98bf7a@news.odyssey.net> ...[color=blue]
> It seems "Andrew Thompson" wrote in
> comp.infosystems.www.authoring.html:[color=green]
> >On 5 Mar 2004 09:49:01 -0800, Safalra wrote:
> >[color=darkred]
> >> http://www.safalra.com/hypertext/html/nojavascript.html[/color]
> >
> >Oh, and you miss-spelled MurkoSoft ..errr
> >MircoSoft, ..err _MicroSoft_ in the last part
> >of the page. ;-)[/color]
>
> I'm looking at the 17:33 GMT version of the page, which shows
> "Microsoft", the correct spelling. (There is no interior capital.)[/color]

I had misspelt it 'Mircosoft'.
[color=blue]
> Your mistaken correction is dated 16 minutes later.
>
> Either Morley has a time machine, or someone's computer time
> settings are wrong, or you are mistaken.[/color]

I have a time machine... Okay, seriously: the last change time on the
page is the time of the last *major* change - it doesn't include
correcting spelling or updating URLs. After I've added a new page, I
read it through again every hour or so, and eventually I find most of
the errors...

--- Safalra (Stephen Morley) ---
http://www.safalra.com/hypertext

On 6 Mar 2004 04:33:28 -0800, Safalra wrote:
[color=blue]
> Stan Brown <the_stan_brown@fastmail.fm> wrote in message news:...[/color]
....[color=blue][color=green]
>> ...Either Morley has a time machine, or someone's computer time
>> settings are wrong, or you are mistaken.[/color]
>
> I have a time machine...[/color]

Thanks for clarifying. ;-)

Say, ..could I borrow it last weekend?

--
Andrew Thompson
* http://www.PhySci.org/ Open-source software suite
* http://www.PhySci.org/codes/ Web & IT Help
* http://www.1point1C.org/ Science & Technology

Andrew Thompson <SeeMySites@www.invalid> wrote in
news:117wognkvzjiu.1g6wat57gcah4$.dlg@40tude.net:
[color=blue]
> On 6 Mar 2004 04:33:28 -0800, Safalra wrote:[color=green]
>>
>> I have a time machine...[/color]
>
> Thanks for clarifying. ;-)
>
> Say, ..could I borrow it last weekend?[/color]

If it existed, you already would have.

It seems "Safalra" wrote in comp.infosystems.www.authoring.html:[color=blue]
>I have a time machine... Okay, seriously: the last change time on the
>page is the time of the last *major* change - it doesn't include
>correcting spelling or updating URLs. After I've added a new page, I
>read it through again every hour or so, and eventually I find most of
>the errors...[/color]

Oh sorry -- I thought "last changed" meant "last changed".
Seriously, I think you might want to reword that footer if you only
record major changes.

Also you might want to fix the arithmetic error (or typo) of 35%
where the correct number is 36%.

--
Stan Brown, Oak Road Systems, Cortland County, New York, USA
http://OakRoadSystems.com/
HTML 4.01 spec: http://www.w3.org/TR/html401/
validator: http://validator.w3.org/
CSS 2 spec: http://www.w3.org/TR/REC-CSS2/
2.1 changes: http://www.w3.org/TR/CSS21/changes.html
validator: http://jigsaw.w3.org/css-validator/

Andrew Thompson <SeeMySites@www.invalid> wrote in message news:<117wognkvzjiu.1g6wat57gcah4$.dlg@40tude.net> ...[color=blue]
> On 6 Mar 2004 04:33:28 -0800, Safalra wrote:[color=green]
> > Stan Brown <the_stan_brown@fastmail.fm> wrote in message news:...[/color]
> ...[color=green][color=darkred]
> >> ...Either Morley has a time machine, or someone's computer time
> >> settings are wrong, or you are mistaken.[/color]
> >
> > I have a time machine...[/color]
>
> Thanks for clarifying. ;-)
>
> Say, ..could I borrow it last weekend?[/color]

I'm afraid I'm already going to book it out - tomorrow I asked myself
yesterday whether it would be free last weekend.

-- Safalra (Stephen Morley) --
http://www.safalra.com/science

Stan Brown <the_stan_brown@fastmail.fm> wrote in message news:<MPG.1ab3c3b9af4fb57b98bf92@news.odyssey.net> ...[color=blue]
> It seems "Safalra" wrote in comp.infosystems.www.authoring.html:[color=green]
> >I have a time machine... Okay, seriously: the last change time on the
> >page is the time of the last *major* change - it doesn't include
> >correcting spelling or updating URLs. After I've added a new page, I
> >read it through again every hour or so, and eventually I find most of
> >the errors...[/color]
>
> Oh sorry -- I thought "last changed" meant "last changed".
> Seriously, I think you might want to reword that footer if you only
> record major changes.[/color]

I'm taking the Platonic view - the document is a flawed image of a
real document that exists in some other place. Spelling corrections
are changes to the image; 'major' changes are changes in which object
the image is a representation of [1]. I'm documenting the later only,
as it is more useful - the last changed date then tells you whether
it's worth reading again because of some major conceptual change since
your last visit.
[color=blue]
> Also you might want to fix the arithmetic error (or typo) of 35%
> where the correct number is 36%.[/color]

Okay, done.

[1] Apologies for the appalling grammar there...

--
Safalra (Stephen Morley)
http://www.safalra.com/hypertext/