SAMPLE SIZE CALCULATOR
Our last issue included a discussion about sampling in survey research.
This issue includes an Excel spreadsheet that you can use to calculate
the appropriate sample size for your research.
Click here to download the spreadsheet. Feel
free to call Bob if you need help at 561.744.5662.
How the spreadsheet works:
There are four inputs used for determining the sample size:
1. Population size
2. Desired confidence interval
3. Desired confidence level
4. Percentage
The population size is entered in column A of the spreadsheet.
The population size represents the number of potential respondents
who match the characteristics desired for the study. For example,
the population might be the number of customers you service in a
specific region, the number of likely voters in your state, or the
number of families with school age children in your district. The
population size can usually be determined or estimated from your
internal databases (for customer surveys) or from census data (for
population surveys).
The confidence interval tells you how closely the results of your
survey represent the entire population. The confidence interval
is the plus or minus figure that usually accompanies a published
study.
The confidence level is how certain you want to be that the results
of your survey will fall within the selected confidence interval.
For example, if you select a confidence interval of plus or minus
4% and a confidence level of 95% then you are 95% confident that
results of survey are within 4 percentage points of the results
you would get if you surveyed the entire population.
Naturally, the ideal is to have the confidence level as high as
possible and the confidence interval as small as possible. However,
interviewing costs and diminished returns from additional sample
can restrict this. For example, with a population of 100,000, we
can achieve 95% confidence that the results are +/-5% with about
380 surveys overall, with no segmentation requirements. To improve
the reliability of our results to 99% +/-5% overall, we need 660
surveys. This represents a significant increase in fielding costs,
without noticeable benefit.
Percentage, the final variable in the sample size calculator, is
the measure of homogeneity in the population. The reliability of
results is greater if 90% of respondents are in agreement, than
if half feel one way and half another. By assuming a percentage
of 50%, we know that all questions fall within the stated confidence
interval at the stated confidence level. Using a percentage other
than 50% occurs only when existing data is available. For example,
if we are doing a customer satisfaction tracking study, the baseline
might show that at least 65% of customers are satisfied with your
performance on every attribute tested. In that case, we can use
60% as a conservative estimate of the percentage.
How to do it – To use the spreadsheet, just type the population
number in column A. The sample size for 95% +/-5% confidence will
appear in column B (you may have to press F9 to recalculate depending
on your settings. You can also change the confidence level, confidence
interval, and percentage as desired. All of the variables are formatted
in Bold Blue, with
the resulting sample size formatted in Bold
Red.
In addition to the sample size calculator, the Excel file includes
a confidence interval worksheet. You can use this to determine the
confidence interval if the survey is already complete and you know
the population size and the number of surveys. The only difference
from the sample size calculator is that you type the sample size
and the spreadsheet returns the confidence interval for the selected
confidence level. ‡
| 
