Determination of the sample size is informed by knowing various factors. These factors comprise of the effect size which entails the difference that exists between two groups under consideration; for continuous data, the standard deviation of the population should be determined. Another factor is the experiment’s desired power for detection of the hypothesized effect and lastly the stated significance level determines the sample size (Dimitris & Aitken, 2009). The last two factors are generally fixed conventionally while the first two are specific to a certain experiment. The enormity of the effect to be detected by the investigator ought to be stated quantitatively, and the standard deviation of the population for the variable of interest ought to be available from the pilot study. The sample size used should reflect the clear picture of the population in question.
Sometimes larger samples may be preferred while in other occasions smaller samples are seen to be sufficient. A larger sample would be preferable in the event that high precision is required. The larger the sample size the higher the precision; this is because when a larger sample is chosen the obtained outcome will give a more accurate representation of the overall population. For a population that shows lack of uniformity, a large sample size would be advisable (Khamis & Radwan, 2010). In some incidences, taking large samples may not be feasible. This may be due to financial constraints and time limitations. When this is the case, then small samples are taken. Additionally, when the population shows high level of uniformity then a small sample size would serve the purpose. If a high significance level is set, then the size of the sample to be collected would be high and when the significance level is small the sample size would be small.
Dimitris, M. & Aitken, C.G.G. (2009). Sample size determination for categorical responses, Journal of forensic sciences, 54(1), p. 135-151
Khamis, N.M. & Radwan, S.S. (2010). Geometric sample size determination in Bayesian analysis, Journal of applied statistics, 37(4), p. 567-575