Hello Zahid,

If the sample size computation is for some target precision of population parameter estimates (e.g., what proportion of Nomadic children in Pakistan attend formal school for at least six years?), then yes, for small finite populations the requisite sample size can be notably smaller than for instances in which the target population is very large or, for practical purposes, infinite.

How much of a difference does it make? Here's some sample values for estimating, via simple random sampling, a population proportion to within +/- .04 (4%) with a desired confidence of 95%, assuming the largest possible variance (0.25 = a 50-50 split in the population), for various population sizes:

Infinite population: N (of cases required) = 601 cases

Pop of 500,000 : N = 601

Pop of 100,000: N = 598

Pop of 50,000: N = 594

Pop of 25,000: N = 587

Pop of 10,000: N = 567

Pop of 5,000: N = 537

Pop of 2,500: N = 485

Pop of 1,000: N = 376

Pop of 750: N = 334

Pop of 500: N = 274

You can always over-estimate the population size if unsure as to the true value. The worst case scenario is that you'll end up with an estimate that has at least, and likely more, precision than what you had originally targeted.

On the other hand, if you can't use a probability sampling method (such as simple random sampling), then the usual formulae won't apply and there's no assurance that your resultant sample will be a reasonable representation of the population.

Good luck with your work.

conduct a pilot study on the population size of 500 and determine the result of how many were in education status . take that as prevalence rate and estimate the sample size for the whole population.

Here is a general approach:

• Identify the confidence level.
• Determine the margin of error.
• Estimate the population standard Deviation.