# Is it necessary to do a pilot study for the sample size calculation for any research?

Most of the researchers do sample size calculation from review and literature. Is it acceptable?

Question

Most of the researchers do sample size calculation from review and literature. Is it acceptable?

- The pros and cons of the arguments are well described and shall become apparent when more discussion occurs, but I have a query with respect to this query: How do we decide the sample size for the pilot study? In many conditions the data obtained from the pilot study maybe skewed by multiple factors confounding the piloting itself (low sample size, clustering of sampled subjects, other biases) leading to confounding of the sample size calculation. In order to weed out the problems of the pilto study sample size we may end up doing a study with too large a sample! How to bypass this problem, then?
- Yes... We must sample size calculation for any pilot works.... E.g: in water and wastewater treatment plant....
- Gentlemen, with due respect, “pilot” study is not needed once you have enough literature to estimate the parameters needed for sample size calculations and power estimation. Pilot study is done solely for the purpose to attain parameters needed for the SS, in case, nature of the study is such that no prior information is available. One important factor which this paper fail to discuss is the effect size which is the major factor driving SS. Due to this reason, collaboration with the experts in the area of the study and a specialist who has good scientific knowledge of statistics are required; filing the blanks and solving a given formula is not the objective here.
- In sample size calculation in epidemiological studies often some knowledge of certain parameters is required. For example, the rate of exposure on the control group. If a reliable estimate of this can be had from previous similar studies then a pilot study is not required. If there is no such estimate then a pilot study has to be carried out. The size of the pilot study should be the minimum size required to get an accurate estimate of the required parameter.
- If your topic is not new, you can use the figures from relevant literatures. It is better to use research with better quality. e.g. RCT will be better than cohort study, cohort study better than cross-sectional study. If possible, systematic review (meta analysis) is always preferable. Also you may find different references to see if the figures are consistent.

If your topic is too new and there is no reference at all, then use pilot study for sample size calculation. Moreover if you are not sure whether the questionnaire is good enough. You need a pilot study. - In most situations, a pilot study is not needed to calculate sample size for a poposed study. The parameters required for the calcuation of sample size MAY be obtained from a pilot study, but this assumes the pilot study is a good, representative sample of the population to be studied, an assumption that is almost always wrong. As mentioned before, a good understanding of the subject matter and evidence from past studies should provide reasonable estimates of parameters, with appropriate margins, to be used in any calculations. Pilot studies would only provide estimates which in itself may be questionable, even with a judicial choice of the sample for pilot study. Having said that, if nothing is known about the parameters of interest, SOMETIMTES, pilot studies may give some clues (but one has to still use judgement and past experience). I find that I am always using a range (I suspect most do) for the estimates of sample size, as obviously there is uncertainty in the parameter estimates we use. It is then a matter of judgement about what number we use for our study, which may be validated only AFTER the study is completed.
- We do not need to conduct a pilot study for calculation of sample size . We can use the information from the previous studies such as effect size, proportion. But if you want to work on a very new topic which there is not any research or even there is not any study that is partially similar to your study you should do a pilot. Of course in some cases you can calculate the maximum size of sample size, but this method is not recommended because increase of resources which you need for your study.

Already a member? Log in

## Popular Answers

Sartaj Alam· Texas Children's Hospital