Edwynn -> RE: Cops and Guns (4/15/2013 8:57:54 AM)
|
quote:
ORIGINAL: Powergamz1 What one learns in the first statistics class is that a truly random sample is impossible, for a variety of reasons from self selection on out. Self selection refers only to voluntary surveys, and non-probability (ergo non-scientific) samples in general. There is little if any self selection in a well designed probability sample, even those involving humans. Biologists use statistics all the time. They use probability samples, the only scientifically valid kind. Bacteria, rats, migrating fowl, etc., do not self select. If one wants to estimate the distribution of empty cola and beer cans around the periphery of a parking lot by counting the cans at randomly chosen locations, those locations having equal probability of being chosen, one can get a reasonably good estimate of distribution of empty cans around that parking lot. The cans do not self select. If OTOH a pollster stops people outside of a superstore and asks where they throw their empty cans (a non-probability sample), the resulting estimate would be quite different, for a variety of reasons. quote:
What one learns in an *advanced* statistics class, is how to control for those factors. Do tell. quote:
In terms of usefulness, n=15,000 is quite acceptable out of a total population of 700,000 If we were talking about a probability sample, where the sample was randomly chosen from the population of concern (all LE members), with each member of the population having the same statistical probability of being chosen, then indeed 15,000 would be overkill. In a non-probability sample (a voluntary survey given to an arbitrary non-random selection of population), the number of respondents is meaningless, especially in this case there being a double bias of selectivity by both parties. quote:
What one learns from real life is how foolish it is to try to make the territory match the map. Which is precisely the purpose of directed purpose well crafted surveys such as this one. Nothing the least bit foolish about it to the marketers. What one learns from chemistry, statistics, history, economics, biology, finance, business, political science, and especially marketing, is that forcing reality, however temporarily and whatever the disturbance elsewhere, match the corporate interest pays damn good money for those really good at it. quote:
And what you might learn from reading the posts above, is that the 400,000 member number includes those who are not LEOs at all. What I learn from that is that the population is narrowed and non-randomly restricted even further than I thought.
|
|
|
|