Tony Cooper
On 18 Apr 2014 11:25:15 GMT, Sandman <mr@sandman.net>wrote:

Sandman
In article <una1l95fqodlib0kd8sbp5vn0pcd9jqtpm@4ax.com>, Eric Stevens wrote:

Which would be statistically worthless even if we were only talking about one store, which we weren't.

Eric Stevens
Why would it be statistically worthless even for the one store at which the data was gathered?

Sandman
Because it is only one data point. It's like I would ask you if you're feeling good and you complain about a headache and then I use this to conclude that you suffer from headaches 100% of the time.

Tony Cooper
All statistics start with one data point. The more data points that are added, the more viable the statistics are. The first set of data points are no more or less significant than the 50th set. It's just that you don't project from a small base with any expectation of reliability.

If I returned to that same Best Buy each week for 49 more weeks, I could make a projection of the number of working units that could be expected to be found at that store, but I couldn't project the figure to other Best Buy locations.

Sandman
You mean, other than his explicit claim about: "I'd estimate that in somewhere around half the visits", which means that he has at least visited enough stores for a number of them being "about half", which puts it roughly in at least around ten stores (4/10 is "about half"), so he already has ten times as much statistical data than you (i.e. infintaely more, since you have none).

Eric Stevens
Do you really call an estimate "statistical data"?

Sandman
Do you really think it can't be?

Eric Stevens
An estimate with some hard numbers behind it, including error data and an acceptable estimate of probability might be regarded as statistical data. Without that kind of support, it's just a guess.

Sandman
No, it's not just a guess. It had some hard data behind it, it's just that nospam couldn't remember the actual numbers, so based on memory, he estimated them to be something close to what they were.

Tony Cooper
He didn't even count or test some of the cameras.