Sampling in Research Series 2: Basic Concepts in Estimating Sample Size

Abstract

Sampling process/ method has been like an examination to the
researchers’ botheration from generation to generation from the
time of inception towards drawing a representative group of units or
cases from a particular population. In series 1 the most popular and
frequently used sampling based on probability theory had been depicted.
In this series, sampling for special types of studies will be discussed.
It has already been emphatically stated earlier ‘no thumb rule of 100’
(or any other fixed number) of samples are permitted for ideal study
or research. Most research publications and MD theses are being seen
to have been accepted in spite of a wrong sample size. Even many
journals today publish papers without proper sample size, although it
is the responsibility of the reviewer who does not know statistics or
epidemiology; the journal concerned is also equally responsible for a
cherry-picking reviewer with partial knowledge. In fact, such journals
should be stopped publishing or be blacklisted since they dispense the
wrong knowledge to the researchers and the young scientist might
learn misguidedly. In most of the population studies, 4pq / L2 where
p=Prevalence; q=100 or 1 - Prevalence; and L = permissible limit of error
has to be <10 % of the prevalence. L or permissible error should not
be taken as absolute term. It’s easy to understand that if the error is
10% admitted before a study then the study stands to be null and void.
There could be more than 100 types of errors while implementing the
study like human error, equipment error, interviewer error, respondent
error and so on. If sampling itself has a huge error of 10%; the study
will become a Pandora’s Box of errors only. Thus, it was imperative
to highlight the value of sample size calculation based on probability
theory to determine the result of chance and types of different studies.