Friday, June 4, 2010

statistics survaying a population

For a 95% Confidence interval
Choose a survey error percent (the percent of deviation from the real result that is acceptable)

InfPopSampleSize = 1/(FractionErr^2)
LimPopSampleSize = InfPopSampleSize * (PopSize /(PopSize + InfPopSampleSize))

Basically samples can be taken any way that doesn't bias the data.

Some to consider are:
* are you limiting your samples to a sub group of the population
* are you dis-proportionally taking samples from a sub group of the population ie

some examples of bad sampling are
- sampling from 1 city for the world population
- taking 100 samples in Tokyo and 100 in Antarctica
- taking samples of people using a web site at 3pm only
- taking samples whole population by using the phone to select them
(ie what about the kids, other members of the family and people who are not listed in the phone book)

No comments:

Post a Comment