Hacker News Reader

Comment by jacquesm | Hacker News Reader

A sample size of '1' and this guy is an expert ? One would hope that Morgan Stanley knows enough statistics to get a more representative sample, or god forbid a panel.

Of course, if they've managed to corner the one 15 year old whose friends are a perfect cross cut of their group then I stand corrected.

If you generalize from such a small sample size to a large number of individuals you are going to have to include some very large error bars around your 'conclusions'.

View on HN

"Without claiming representation or statistical accuracy, his piece provides one of the clearest and most thought provoking insights we have seen. So we published it."

From the actual report, linked by lucifer:

http://media.ft.com/cms/c3852b2e-6f9a-11de-bfc5-00144feabdc0...

Here is the article on Bloomberg: http://www.bloomberg.com/apps/news?pid=20601204&sid=aG2U...

Sample size is the kid + his friends.

If you just choose a child at random, you are likely to get an average child. There isn't just one average kid, there are MANY, and a few weird ones.

This kid is from a certain social background and is likely to associate with kids from a similar social background, unless he is a networker par excellence and has contacts all over the globe with teenagers from every income level and education level.

There are untold variations on that, and each of those subgroups has their own 'averages', this kid is at best - and I'm not saying that he is, just leaving open the possibility - in the middle of his own subgroup.

Of course, but these are a bunch of middle-aged analysts we're talking about. Actually getting some first-hand information from teenagers is going to be a huge step up regardless of statistical bias.

any suggestions for items in a questionnaire? :)

classic interpretations of the Central Limit Theorem considers a sample size of n>29 to be "large"; so if this guy has 29 other friends, that's a "large" enough SS

Reply on news.ycombinator.com