Random Sampling Bias

Six Sigma – iSixSigma Forums Old Forums General Random Sampling Bias

Viewing 10 posts - 1 through 10 (of 10 total)
  • Author
  • #26908


    Welcome. We hope you enjoyed the article ( and also hope that it spurred on some thoughts about the way sampling is performed at your business or organization.

    We wanted to discuss a follow-up subject having to do with the ebusiness example. You should always be careful when conducting surveys of people, because bias is always an issue. People tend to give their opinion most when their experience falls at the extremes — when they either completely love something or completely hate something. If this is the case in your survey, those in the middle would probably not fill out the application and the data resulting may be biased.

    How can you get around this issue? Let’s hear your thoughts!

    The Quality Internet



    You could always entice the user to give his/her reply regardless of the “passion” behind it. Maybe you could give away a gift certificate or one of your products or something?


    John El

    Money. Promise that you’ll send them money if they fill it out and provide an email address. That way if they do a poor job you can withhold the cash!


    Michael Rusignola

    We were just having this very discussion at work today! We were attempting to determine whether a situation similar to the e-business example required a true random sampling of data to be valuable.

    My argument is that since we cannot assert that every individual is equally likely to take the optional survey, the data is thereby derived from a non-random selection pool. This makes it impossible derive any generalizations about your populous. You may, for example, prematurely decide to end an ad campaign with Yahoo! unaware that the demographic that visits Yahoo! is statistically less likely to fill out online marketing surveys.

    The other person’s point-of-view was that any data, whether skewed or not, is valuable. Marketing is an imprecise science and many things are determined in a “hit-or-miss” fashion. Whether or not the data is mined from a truly random sample is irrelevant. The information received is valuable regardless of the manner in which it was obtained.

    If it can be determined, who is right: the statistician or the marketer?


    Ken Myers


    As you already know survey accuracy can only be achieved through an unbiased sampling of the site users. If the users who answered your survey come from limited demographics, then the response will be representative of that demographic only.

    Decisions made from a biased survey can be just as poor as not using a survey as a reference. If your management or marketing group intends on using this survey data, someone should ask what the difference is between the respondent demographics and the intended demographics. If this difference is not great, then the survey may be of practical use. Otherwise, the survey could prompt considerable decision making error.



    I guess I would fall somewhere in the middle of the two opinions expressed here. I think you can probably learn something from surveys, even if you can’t control the randomness…but to say that the sampling is irrelevant is dangerous. Always be aware of the limitations of your results.

    Let me give an example that is real world, but not related to marketing. In Psychology (yes…this Six Sigma-ite is a Psych major — gasp!), most studies done in university settings use hapless freshman Psych 100 students as the subjects for study. Many times a requirement of their class is to particpate in a specified number of research studies during the semester. My senior project used 20 such freshman for my study. I did EEG’s (brain waves) on them while they were completing a verbal and spatial task (to test the left brain/right brain theory). Was my sample representative of the human population. I’ll stick my neck out here and say NO!!! I had 20 middle to upper class 18 year olds who weren’t too happy about being there. Was the information still valuable? Definitely. When we wrote up the results, it was important to discuss the limitations of the study based on sampling. The name of the game was replicate replicate replicate. Had we gotten very conclusive results (we didn’t), the study would have been repeated under more controlled circumstances, at least where sampling was concerned.

    Suggesting giving monetary or other rewards for survey participation always makes me nervous. It seems that this adds another possbility for skewing the results…trying to please the adminsterer of the test to better the chances of getting a reward.



    good day!  how will i get the sample from a public place like libraries? people come and go to those places.  thanks!



    Donna, Good evening!About your question regarding a ” … sample from a public place like libraries … ” — your request is a little short on specifics.What kind of sample are you trying to get? Is this an opinion survey? And, why from a public place?
    Best regards,



    Hi everyone…I needed your expert opinion on one of the important matters here in my biscuit manufacturing industry…
    the speed of machine is 2500 biscuits/minute in a total of 16 parallel lines and a shift is 8 hours long…
    so what should be my sampling frequency as well as sample size..right now i take 16 samples, at a gap of 30 minutes with each sample made of 32 biscuits(2 from each line)………is it right…
    will be grateful if you could help me out



    There is no right or wrong.  There is only the amount of risk your sampling plan generates and the amount you are willing to live with (or what your market will bear). 
    It sounds to me like you have a continuous manufacturing process.  If that is the case, you should probably be using a continuous sampling plan.  Try doing some research on them (i.e. use google) and see if it works for you.

Viewing 10 posts - 1 through 10 (of 10 total)

The forum ‘General’ is closed to new topics and replies.