Talk:Sampling (statistics)
From Wikipedia, the free encyclopedia
A question that's been bothering me for some time: how does population size affect sample size for the same probability of precision? (I hope I asked this correctly.) Wblakesx
The proportion of sample size to population size is called the sampling fraction. Generally speaking, the precision depends on the absolute number of samples, taken and NOT the sampling fraction. Thus a sample of random of 1000 people is fine for estimating the views of a population of 60 million (as in the UK) but it would be almost equally good for estimating the views of a population of 600 million of 6 billion. Having said that, it's not quite true. As the sampling fraction becomes significant (greater than 5 or 10%, say) you do need to add a correction to your precision estimates. But it's a good question, because the layman's intuition that the sampling fraction must be large for the sample to be good is dead wrong! Blaise 22:09, 20 September 2005 (UTC)
What is random sampling ?
The first line of the Mechanical Sampling section reads:
Mechanical sampling does not occurs typically in sampling solids, liquids and gases, using devices such as grabs, scoops, thief probes, the coliwasa and riffle splitter.
This doesn't make any sense. I don't know anything about mechanical sampling so could someone who does please fix this? TooMuchMath 03:06, 15 May 2006 (UTC)
Contents |
[edit] Indian invention of sampling in 1928
User:144.92.82.172 added this statement to the article: "Some say that sampling was invented in India in July 1928." Can anyone provide a source explaining this claim? I have removed it from the article for now. -- Avenue 13:17, 25 May 2006 (UTC)
- This claim does not proceed; Sampling is a intuitive concept and happens (by instinct) since the beggining of time. Also, there are records of use of previously planned samples since the XIX century.--Lucas Gallindo 20:06, 29 September 2006 (UTC)
[edit] Section on case studies and sampling
Editors Krakenflies and Gsaup recently added this extensive section on "sampling" in case studies. Although I enjoyed reading the linked article, I believe the section does not relate to the topic of this article, statistical sampling, so I will delete it. Perhaps it could go in a new Sampling (case studies) article instead. -- Avenue 11:51, 11 August 2006 (UTC)
- Or maybe combined with Theoretical sampling? -- Avenue 12:09, 11 August 2006 (UTC)
[edit] Added definitions for types of data and levels of measurement.
I thought these subheadings would be relevant. Let me know. JT Pickering 22:20, 3 September 2006 (UTC)joe pickering
[edit] Outdented 'levels of measurement'
Seemed to flow better when not considered as part of the types of data.JT Pickering 18:29, 12 November 2006 (UTC)
[edit] Sampling used as Democratic Process
I would like to know of any movements, if they exist or ever have existed, which advocate for the use of random sampling techniques to provide official decision making bodies. For example, what if the US House of Representatives were actually a representative sample? This seems like something which could be worthy of exploration since it seems one of the only ways to remove bias toward the rich and the non-competitive climate would almost certainly be more conducive to productive intellectual dialogues, as opposed to degenerate partisan contests. Maybe then politics wouldn't be such a turn off to so many people. LordBrain 23:21, 6 October 2006 (UTC)
[edit] Stratified sampling
At the end of the Stratified sampling section it says:
"Typically, strata should be chosen to have:
* means which differ substantially from one another * variances which are different from one another, and lower than the overall variance."
Wouldn't the second bullet be more clearly expressed as "minimise variance within strata and maximise variance between strata."?
It might also be worth mentioning that stratified sampling can introduce bias when selecting strata.
