Re: Small Sample Size and Validity



nicgushar@xxxxxxxxx wrote:
How does reducing sample size effect validity?

(I take it that you meant "affect" ?)

I can't explain mathematically but I can explain more logically.

Each member of a sample contributes something to the validity of a result. Furthermore, there is a probability that each member of the sample is wrong, which is more or less independent of the number of samples (it might be dependent on the distribution of samples, however). Given those two pieces of estimates, for small sample sizes the probability that the sample is wrong begins dominate its contribution to the validity.

An easier way to put it is that the smaller the sample size, the more correlated the data needs to be to be valid.
.



Relevant Pages

  • Re: Presumed guilty
    ... No one decides "balance of probability" so, ... No student would ever be allowed to be alone, unobserved, with a member ... if other members of staff had been in a position to state ...
    (uk.legal)
  • Re: Continuity of spacetime
    ... probability of existence of any given x_n is %100. ... No. I'm giving the reals a new property, ... validity will always be indeterminate. ...
    (sci.math)
  • Re: OT: Everyone born after a certain date is ignorant and can be dismissed
    ... the sample the less likely any member of the sample is to skew the result. ... admit that my example using ten was not inherently improbable ... If increasing the sample increased the probability from 5%, then, logically, ...
    (uk.sport.cricket)
  • Re: Can Events of Zero Probability Happen?
    ... What is the probability that X = 1/2? ... the mathematician is just as stuck as the physicist. ... If one has a set which is countably infinite, like the set of integers, ... "chose one member at random", since that would require each of countably ...
    (sci.math)
  • Clustering Probabilities
    ... Given that a known n number of clusters are required, and the probability that each member of a data set belongs to each of the n regions. ...
    (sci.stat.math)