Will changing sample size also change our awareness
results in our consumer survey?  How exactly did you decide
on our sample size?

Response:

Changing the number of respondents will not cause awareness to inflate or deflate, as long as the area surveyed remains the same.  For example, whether we survey 100 or 1000 consumers in zip code A, there would be no impact on the results, other than the results would be more reliable with the larger sample size. Awareness results are shown as a percentage, so as long as the sample size is within a reliable error range changing the sample size will not shift the results.

Recommendations for sample size are primarily based on error range.   In HealthStream Research’s experience from more than a million surveys, our experts have found that a sample size of 300 to 500 provides the best combination of statistical reliability and project investment for consumer studies.  “The best specific number for error range is a myth.  It’s more like cooking instructions for a microwave dinner that say to cook on high for 6 to 8 minutes, or maybe oven instructions to cook until ‘golden brown.’  You won’t get statisticians to agree on a specific number.  For consumer studies, we typically recommend a sample size of 300, with at least 200 coming from the primary service area,” said Berke Bilbay, Vice President of Data Analysis and Reporting.  “This keeps the error range well within comfortable limits.”

At the 95% level of confidence, the maximum expected error range for a sample of 300 respondents would be ±5.7%.  In other words, if 300 samples of 300 consumers each were randomly chosen from the entire population of a defined service area, 95 times out of 100 the total results obtained would vary by no more than ±5.7 percentage points from the results that would be obtained if all consumers were surveyed. 

“There comes a point of diminishing returns, though. A sample of 100 gives an error range of ±9.8%, while just adding an extra 200 to the sample size causes the error range to drop more than 4 percentage points to ±5.7.  But to improve the error range by an additional 4 percentage points, from ±5.7 to ±1.7, you would have to increase the sample size to a budget-busting and time-wasting 3,200.  A sample size of 300 generally provides the best mix of statistical validity and value,” said Bilbay.   

However, adding more zip codes may cause the awareness number to deflate.  For example, if we survey only consumers living very close to a particular hospital, awareness would likely be much higher than if we added zip codes from a wider area.  The farther away from a facility, the lower the awareness tends to be. 

Do you have a question you’d like to send to our team of experts?  Send an email to researchinfo@healthstream.com, and our interdisciplinary team of experts will provide you with a detailed answer.  Please specify if you wish to remain anonymous.