Surveys: Number of Responses

[Nadeem A.]

Number of responses

Hi,

What should be the minimum number of responses to analyze any kind of survey statistically? Somebody told me 25, Is there any formula associated to this number or it is fixed?

Your reply with reference will be highly appreciated?

Regards,

 

[D.Scott]

Automotive recommends a minimum of 25 subgroups with at least 100 samples.

Be sure though that the process is stable before taking your sub-sets.

Dave

 

[Nadeem A.]

Thanks Dave for the response. Let me make it clear that I am going to analyze some hypotheses related to the ISO 9001:2000. In order to do that research, a survey sheet has been distributed to 138 companies in Quebec and Ontario. We have got 26 responses so far through mail.

Should I start analyzing the responses as I have already got the minimum number of required responses (n=25)?

I am not a statistics guru so that's why I wanted to ensure from any stats champion.

 

[D.Scott]

Hmmmm, I guess I misunderstood the original question. It would be my experience from my old promotional advertising days that a 19 % response to a survey is fantastic. I would do your analysis of that survey whenever you are ready with little expectation of receiving more responses.

I guess I don't see how a statistical analysis of the answers you receive will be valid unless you are going to simply use the % per responses. Statistical analysis as I am used to it in automotive processes just doesn't appear to fit. Unless of course I am missing something - which happens quite often.

Dave

 

[Darius]

Nadeem

I don't know for the Automotive industry but as far as I know there is not a constant for the number of responses

The two factors that determine the sample size of a survey are:

* the confidence level
* the precision (or reliability) range.

I added a Handbook, read page 24-28 of the original document could be of any help.

darius.

 

[Graeme]

There are methods to analyze small samples (number of responses) from different kinds of surveys. After all, that is what the large polling organizations (Gallup, Harris, Nielsen & etc.) do for a living. The drawback of having a small sample size is that your confidence interval (uncertainty) is larger than what is normally seen in an industrial environment.

The only applicable reference I have on my desk right now is Handbook of Statistical Methods for Engineers and Scientists (Harrison M. Wadsworth, 1990.) In particular, see chapter 9, "Survey Sampling Methods."

Other useful references for the type of methods you need would be those that deal with statistical analyses in medical research or the social sciences, or other fields where small samples are the norm. In an ISO 9000 context, there are several publications on measuring customer satisfaction that may have useful methods.

 

[Nadeem A.]

Thankyou very much to all of you for comments and help. I will search the attachment file and the Hand book of Statitistical Analysis for Scientist and Engineers to initiate my work in the right direction.

Also, I am going for the next step of calling companies to send us back the survey sheet. That will help in increasing the responses bit more.

Thanks Dave for your help in the analysis. I am working on it and will share on this forum.