How many people to interview?
Six Sigma – iSixSigma › Forums › Old Forums › General › How many people to interview?
- This topic has 14 replies, 10 voices, and was last updated 14 years, 3 months ago by
Eric B..
-
AuthorPosts
-
February 28, 2008 at 7:40 am #49462
ie96axadParticipant@ie96axadInclude @ie96axad in your post and this person will
be notified via email.I work at a company with 90 employees. We are planning to conduct an interview survey on a sample of the total population. How many people should we interview to get a statistically reliable sample?
0February 28, 2008 at 6:11 pm #169096
Heebeegeebee BBParticipant@Heebeegeebee-BBInclude @Heebeegeebee-BB in your post and this person will
be notified via email.Start your journey here:
https://www.isixsigma.com/st/data/0February 28, 2008 at 7:27 pm #169103ie96axad:When you have such a small population, you need to consider a correction to the usual calculation for sample size. Search for information on “hypergeometric sampling small sample size”.An example to get you started is:http://uregina.ca/~morrisev/Sociology/Sampling%20from%20small%20populations.htmCheers, Alastair
0February 28, 2008 at 8:13 pm #169105To answer your question:
What type of questions are you asking? (True/False, 1-5 scale, etc)
What is your desired confidence?0February 28, 2008 at 8:17 pm #169106To answer your question:
What type of questions are you asking? (True/False, 1-5 scale, etc)
What is your desired confidence?0February 28, 2008 at 8:34 pm #169111How are the subjects being selected? Randomly? If not, the selection process itself would probably introduce bias into your results.Out of curiosity, how vital is it that the survey results are “statistically reliable?”
0February 28, 2008 at 8:46 pm #169114
TaylorParticipant@Chad-VaderInclude @Chad-Vader in your post and this person will
be notified via email.for such a small group, why not do them.
but i did see in another post something about 42
0February 28, 2008 at 9:51 pm #169115Your sample size will depend on the variation in the data for each question asked. Hence, each question will probably have a wildly different sample size requirement, and most will probably require many more than your population. Within the same survey I have seen some questions require as low as 40 and as high as 2500.So seeking a sample size as a subset of 90 may be theoretically possible on some questions, but is unlikely on most. In such cases I find it best to survey the entire population.
0February 29, 2008 at 3:32 pm #169138Here is an article I came across on sample size calcs, which includes an excellent coverage on surveys, especially Likert based surveys.http://www.surveyz.com/client/sampling.doc
0March 3, 2008 at 8:19 am #169185
ie96axadParticipant@ie96axadInclude @ie96axad in your post and this person will
be notified via email.I´m asking people to estimate how many hours they spend on a certain activity per week, and I´m looking for 95% confidence. Based on the links posted in this discussion forum, it seems I need to interview all persons in the total population…
0March 3, 2008 at 8:22 am #169186
ie96axadParticipant@ie96axadInclude @ie96axad in your post and this person will
be notified via email.The persons are selected based on time at the company, since at least 1 year employment is needed to be able to answer the questions we are asking. Since this survey might result in head count reduction out of the total population, we need to be able to show reliable nr to support this – that´s why it needs to be statistically reliable
0March 6, 2008 at 2:21 pm #169308Treat responses much like you would a focus group. Whatever responses you get are indicative of feelings that exist among the employees. Mgt can then use this information to make a more informed business decison. Also, if your company has, say admin, production, programming, etc. look at responses by group to eval appropriate repsentation.
0March 6, 2008 at 5:40 pm #169314
George ChynowethParticipant@george-chynowethInclude @george-chynoweth in your post and this person will
be notified via email.With a 95% confidence interval and 3% error, you need to sample 83 people.Sampling is done when the population size is large and unwieldy. It is usually a cost/benefit kind of decision. In your situation it looks like it would be more costly (in terms of time and data quality) to select a sample. Simply survey all of them.
0March 6, 2008 at 9:54 pm #169321
Eric B.Participant@Eric-B.Include @Eric-B. in your post and this person will
be notified via email.I agree with George. You’re going to have to survey nearly all of them to get the level of confidence you want anyway. Ask them all. Eric.
0March 6, 2008 at 9:54 pm #169322
Eric B.Participant@Eric-B.Include @Eric-B. in your post and this person will
be notified via email.I agree with George. You’re going to have to survey nearly all of them to get the level of confidence you want anyway. Ask them all. Eric.
0 -
AuthorPosts
The forum ‘General’ is closed to new topics and replies.