Comment of the Day - November 1, 2005

I guess if you sound confident, that makes it true.

The good part is IE is losing share, and better browsers are gaining.

One of the reason for the FF drop in August was by some attributed to start of College and students having new computers with Win IE default. They now have had time to change this back to firefox or whatever they liked.

Maybe that accounts for the FF rebound.

I also wonder how the release of the free Opera has affected this. Another browser that is much better than IE and IMO FF too.

bstadil - lookup dinotrac's comments beneath my original comment, he made the same proposal. Moreover, earlier he made the same proposal, but that time to explain the microscopic change in FF supposed usage downward. [You could argue great minds think alike, but be careful: I am known to carry a sharp pin.] Nonetheless, without a good survey, which specifically ask those sort of questions that are then compiled and extrapolated it is no more than a educated guess.

My contention is that without explicit statements about how this outfit obtained their raw data and then interpreted it is essentially meaningless. Moreover, they completely seem ignorant of the meaning of the term "significant figures". At best they can assert they are measuring real time usage on a subset of web sites interpreted on some basis. Without those clarifications, just about everything claimed is fraudulent.

txt --

Fradulant is a bit strong, but meaningless is not.

It's funny how little people understand the meaning of "significant" when dealing with data that is statistical in nature.

It all boils down to one word: random. If you wish to use a subset of the universe to represent the universe as a whole, you need to know that:

a) The universe is completely uniform and it doesn't matter which subset you choose because all subsets are essentially identical, or

b) That the members of your subset have been chosen randomly from the unrestricted universe, and

c) that your sample is sufficiently large in light of variation between potential samples.

No magic, no mirrors. Very straightforward. If you can't pass a or b and c, you don't get to claim d.

dino - I am sticking with "Fraudulant" with this outfit, due to their big fuss with less than a percentage point perceived change. In general an overstatement, however, you are not correct this time.

Next I usually cut parts out of my posts before they hit the thread, I had a paragraph discussing some of the factors that should have been discussed by the authors. In this case, I am glad I left it out:

1. you did it better!

2. concise and clear.

3. lots of white space

4. good code, good conclusion

[No sarcasm intended nor present, despite my earned reputation in that regard.]

very well done,

txt

txt --

Aw shucks!!

Thanks.

;0)