Re: Two CSS Classes. Is this possible?
- From: Harlan Messinger <hmessinger.removethis@xxxxxxxxxxx>
- Date: Wed, 24 Sep 2008 13:10:46 -0400
Jukka K. Korpela wrote:
Harlan Messinger wrote:
Lies, blatant lies, statistics, Internet statistics - do you really
want to go that way?
If, for example, the statistics for a government website that receives
over 40,000 visits a day show that 28% of the requests from IE are
currently from IE6 (as is the case), the actual usage may be 35% or
25% or 20%
Or something else. Moreover, which "actual usage"? Usage when accessing that site, or web access in general? One site's statistics, even if it were meaningful in its own context, says nothing about another site's usage.
By the way, according to most statistics I've seen, IE 6 is still more common than IE 7, or about equally common. Anything older than IE 6 is almost ignorable by now, which is of course good news to authors.
So the
information is useful as long as one is aware of the margin of error
involved.
Which margin of error? You don't know it. You _might_ have a reasonable good estimate of the margin error for a _site_ (that is, you might be able to say that with reasonable probability, visits to a particular site are made with, say, IE 6 in 40 - 60 % of cases. But regarding web usage in general, what would you base the estimates on? Gut feeling? Then please don't use pseudostatistical terms like "margin of error".
Absent any particular reason to think that the millions of otherwise diverse people with IE6 have substantially different browsing habits from the millions of otherwise diverse people with IE7, it's a reasonably good estimate for the population at large. That isn't a gut feeling, it's a statistically sound observation. And so I chose to use a pedestrian term instead of "confidence interval" or "standard error". Got anything else pointless to pick on while you're at it? Would you like to challenge me on degrees of freedom and Bayesian analysis while you're at it?
.
- Follow-Ups:
- Re: Two CSS Classes. Is this possible?
- From: Jukka K. Korpela
- Re: Two CSS Classes. Is this possible?
- References:
- Two CSS Classes. Is this possible?
- From: shapper
- Re: Two CSS Classes. Is this possible?
- From: Jukka K. Korpela
- Re: Two CSS Classes. Is this possible?
- From: shapper
- Re: Two CSS Classes. Is this possible?
- From: Jukka K. Korpela
- Re: Two CSS Classes. Is this possible?
- From: Harlan Messinger
- Re: Two CSS Classes. Is this possible?
- From: Jukka K. Korpela
- Two CSS Classes. Is this possible?
- Prev by Date: Re: Two CSS Classes. Is this possible?
- Next by Date: Re: Two CSS Classes. Is this possible?
- Previous by thread: Re: Two CSS Classes. Is this possible?
- Next by thread: Re: Two CSS Classes. Is this possible?
- Index(es):
Relevant Pages
|