Click here to go back to the main website. Start a Live Help session.

## Knowledge Base » SigmaXL Toolbox

Using Sample Size Calculators

on January 01, 2012

How do I use the Sample Size Calculators?

Using Sample Size Calculators

SigmaXL's sample size calculators are based on a confidence interval approach, where you enter the desired half-interval value (delta). For example; in a survey where the outcome answers are yes/no or Republican/Democrat, you would use the Discrete Calculator. Let’s say that you desire the proportion margin of error to be +/- 3%. In this case, you would enter a delta (half-interval) value of .03.

If the survey answers were on a continuous scale of 1 to 5, and you desired a margin of error on the mean to be +/- .25, then you would use the Continuous Calculator with .25 as the delta value.

Note that in this continuous case the units are not percentages but a metric like level of satisfaction.

The challenge with these sample size calculators is that  an estimate of the population proportion for discrete or an estimate of the standard deviation for continuous is needed.  This is a bit of a chicken and egg situation - which comes first? You want to determine an appropriate sample size to determine an outcome, but you are being asked to enter an estimate of the population proportion or standard deviation.

Keep in mind that this tool is a planning tool, the true confidence intervals will be determined after you collect your data.

In the discrete case, estimating the population proportion can be done if you have good historical data to draw from, for example; historical customer surveys where the percentage of satisfied customers was 80%. In this case you could use P=0.8. If you do not have a priori knowledge, then leave P=0.5 which gives the most conservative value (i.e. largest estimate of sample size).  If you do enter a value other than 0.5 it will result in a smaller sample size requirement.

Note that when Gallop is planning surveys, they always use p=0.5.

It’s a bit trickier for continuous data where you need an estimate of the population standard deviation.  If you have historical data you could use an estimate from that.  If you have no idea what the standard deviation will be, then you could take a small sample to get a rough estimate of the standard deviation.