Non-availability of sound guidelines for sample size estimation is the primary factor affecting the quality of extension research in the country. In this blog, Dr P Sethuraman Sivakumar presents guidelines for choosing adequate sample for extension research.
P. Sethuraman Sivakumar is Senior Scientist, Central Tuber Crops Research Institute (ICAR-CTCRI) Sreekariyam, Thiruvananthapuram – 695017, Kerala, India.
Email: sethu_73@fulbrightmail.org
I congratulate Dr Sethuraman for his blog, which is of high standard and of academic brilliance. The determination of sample size in research is of critical concern, which is well covered in the blog To be frank, I have not fully understood some of the concepts (sourcing sample) and beg to differ with some of the ideas given ( I am of the opinion that sampling methods mostly determine the sample size). I also doubt as to how sample size can be determined based on choice of statistical analytical procedure? I feel researchers decide which analysis has to be done after collection and tabulation of data. Illustrative examples would have enhanced the readability of the blog My views are as below: – Sample size problems vary widely in their complexity The most important aspect is that it is context dependent. – Sample size is one aspect of study design. To help determine sample size, many questions have to be answered. Some of the questions are: – Objectives? – Response variable? How measured? – Estimate of non response rate – Sources of variation? -Whether normally distributed data can be ensured? Some other related points to be considered are: – desired width of a confidence interval – prior distribution in a Bayesian context – precision of estimation desired – pilot study – relevance of historical data? -effect size measures described by Cohen- how much dependable?
Sir Thank you for your critical and constructive comments. I will agree with you that sample size determination depends on several factors like sampling method, nature and size of the population and choice of statistical analytical methods. I believe that sampling methods like probabilistic or non-prob samples explain the ways eliminating BIAS (systematic error) and SAMPLING ERROR (random error) in selecting the samples, but not the adequacy of sample. Sampling adequacy or adequate sample is a statistical sample size that is large enough to provide required precision of the survey or test results by minimizing the effect of chance. In scientific research we are trying to explain a phenomenon by describing it, explaining relationship between its components and with other phenomenon with an objective to enhance our ability to replicate it in a similar context. Statistics is the major tool which aid us to perform this description and explanation in a objective way. In this work, the sample size estimation is viewed in terms of accuracy of end results with respect to population parameter, which is a essential part of generalization of research results. Our extension professionals view statistics as a SUFFIX component which emerges after collection of data. As I said earlier, statistics is a major tool which helps to describe and explain a phenomenon. In any scientific research, this description or explanation is a Research purpose which is expressed in objectives (e.g. to assess, explain, describe, predict etc). It is essential for the researcher to know how systematically assess or describe the phenomenon through appropriate analytical strategies (statistical methods). The research instrument development phase, we are dealing with reliability and validity, which are essentially analytical indicators (e.g. correlation, regression, factor analysis). In scientific research, we are dealing with statistics at every phase problem identification (e.g. research prioritization using statistics), objective formulation (knowledge of appropriate statistics for analysis to describe, explain, predict), sampling (sample size using statistics), research instrument development (reliability and validity using statistics), data coding (missing data, outliers etc) and data analysis.
I congratulate Dr.Sethuraman Sivakumar for coming out with a very useful blog on sample size in social research studies. Many a time, due to flaws in sample size and sampling plan which restrict the use of appropriate statistical tools fails to come out with quality research articles. Consequently, we cannot submit such articles ( poor quality ) to good journals and look for mediocre journals. It is high time that extension researchers must focus on quality research. In this context this blog is very comprehensive with good examples and certainly helps all those who are interested in good quality research.
Sample size selection is always a problem for extension researchers, especially for PG and PhD Scholars. Many are following the rules of thumb from unknown sources. Thanks for providing the guidelines. Also, it will be good to have list of statistical tools and techniques, which could be used for small samples
This blog is an attempt to put relevant guidelines in one place to help the budding scientists and students in selecting adequate samples. Sample size is one of the primary factor which determines generalisability. Many studies demonstrated that small sample research works produced inconsistencies when replicated in similar systems. In general, small samples are advisable only when the research is conducted in a distinct and unique system (e.g. geographically-inaccessible, high disaster prone areas) and the relationships with large effect size.