|
|
Is there a Response Rate Fallacy (if not there should be IMHO)Assuming that the commonly published figures are correct - internal surveys will generally receive a 30-40% response rate (or more) on average, compared to an average 10-15% response rate for external surveys - we are left with a 70 - 85% non-responder rate. There are significantly lower response rates for online surveys (which now are the norm) when compared to paper based surveys. How can we draw any concrete conclusions based on the usual response rates due to non-response bias and the like? To do so would be something of a Response Rate Fallacy. i.e. it is a fallacy to derive much meaning from a survey with a low response rate unless it is assured that the sample was representative and even then one is on thin ice.
|
| asked on Saturday, Apr 01, 2017 12:45:58 PM by JK | |
Top Categories Suggested by Community |
|
Comments |
|
|
| |
Want to get notified of all questions as they are asked? Update your mail preferences and turn on "Instant Notification."
Part one is about how science works even when the public thinks it doesn't. Part two will certainly ruffle some feathers by offering a reason- and science-based perspective on issues where political correctness has gone awry. Part three provides some data-driven advice for your health and well-being. Part four looks at human behavior and how we can better navigate our social worlds. In part five we put on our skeptical goggles and critically examine a few commonly-held beliefs. In the final section, we look at a few ways how we all can make the world a better place.
* This is for the author's bookstore only. Applies to autographed hardcover, audiobook, and ebook.
|
|
The response rate is irrelevant to the research. As you alluded to, the sample size being adequate is the real consideration. For example, if survey 1 received a 1% response rate and was sent to a total representative sample of 10,000 people, it would get 100 responses. If survey 2 received a 10% response rate and went to the representative sample of 1000 people, it would get 100 responses. Assuming no shenanigans, the results from both surveys should be similar. The statistical problem with poor response rates arise when the sample is not large enough to meet the desired confidence level based on the anticipated number of responses.
If I am missing something here, let me know. |
| answered on Saturday, Apr 01, 2017 02:44:13 PM by Bo Bennett, PhD |
Comments |
|
| |
