Linked by Thom Holwerda on Mon 2nd Feb 2009 12:00 UTC
In the News Since it's nigh on impossible to produce accurate figures of operating system usage, we have to make do with figures that provide a rough estimate, at best. One such set of statistics are the figures from Net Applications, which tracks the 160 million monthly visitors to its hosted websites. The latest figures from January 2009 have been published, and they show that the rise of Mac OS X continues, as well as that of the iPhone and iPod Touch. Unsurprisingly, Windows 7 did quite well too.
Thread beginning with comment 346622
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE: Sample Size
by alcibiades on Mon 2nd Feb 2009 15:32 UTC in reply to "Sample Size"
alcibiades
Member since:
2005-10-12

is that 160 monthly visitors the correct sample size and not a typo? Because if it is correct, 160 is not an adequate sample size for something like this.


Only if it is truly random, in which case it would be ok. Its the randomness that is critical, raising the numbers will not help if it is non-random.

Non random sample reports are amusing starts to a conversation, but they are not data.

Reply Parent Score: 2

RE[2]: Sample Size
by GODhack on Mon 2nd Feb 2009 15:43 in reply to "RE: Sample Size"
GODhack Member since:
2008-05-16

Yes but 160 means some fuuny things:
And HOW FreeBSD came up with 0.01. One Free BSD user come up there and generated those nubers. Maybe there is one lucky FreeBSD user in entire world?

I had statistic (biometrics) lectures in university and based on them I can say that these numbers mean NOTHING.

Reply Parent Score: 3

RE[2]: Sample Size
by pantheraleo on Mon 2nd Feb 2009 17:26 in reply to "RE: Sample Size"
pantheraleo Member since:
2007-03-07

> Only if it is truly random, in which case it would
> be ok. Its the randomness that is critical...

The sample size is also important. If your sample size is too small, you won't be able to get a meaningful P value, and this means you cannot determine whether the change is statistically significant.

Edited 2009-02-02 17:29 UTC

Reply Parent Score: 1

RE[3]: Sample Size
by GODhack on Mon 2nd Feb 2009 17:49 in reply to "RE[2]: Sample Size"
GODhack Member since:
2008-05-16

And from other side if sample size is 160 000 000 you will not be able even to input this number into computer statistics programs.

And if you do so huge sample size it means you are stupid and wasting time and money.

Statistics invented in XIX century just because we need method to find information about huge population in fast and cheap way. I looks like some big companies find XIX century math too difficult for them even in computer age.

Reply Parent Score: -1