Review data from pilot studies. Review historical data from completed studies with similar characteristics in the absence of data from pilot studies. Previous studies are good sources of statistical information. Remember, sample size is the required number of observations in a sample. Sample size is estimated from the standard deviation or expected proportion, width of the confidence interval and the confidence level.
Calculate the variance for data from a pilot study. Remember that variables could be continuous, discrete or categorical while calculating variance. Estimate the standard deviation by calculating the square root of the variance. Use standard deviation from a previous study in the absence of data from a pilot study. The standard deviation from a previous study on obesity in children aged 2 to 5 years, for example, was found to be 15 lbs. Estimate the proportion of observations with one of the values in the case of dichotomous variable. Use the proportion without the value if the estimated proportion is more than 50 percent.
Specify the width for the range of values that contains the true value. The estimate of confidence interval is the margin of error. We want to determine the mean weight of children aged 2 to 5 years, for example, with a confidence interval of 5 lbs. The width of the confidence interval is 10 in this case.
Select a confidence level for the confidence interval. A 95 percent confidence level, for example, means that we're 95 percent certain that the interval includes the true value. Remember, a narrow confidence interval with high confidence level has a higher probability of including the true population value. We want a 90 percent confidence level, for example, for the study.
Calculate the sample size from the standard deviation or expected proportion, width of the confidence interval and the confidence level. The estimated sample size is 25 for the study with standard deviation of 15, confidence interval of 10, and 90 percent confidence level.