Suvey Data and MSA

Six Sigma – iSixSigma Forums Old Forums General Suvey Data and MSA

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
  • #33317


    Hello – One of my projects is to improve the new hire process.  We collected cycle time and yes/no data around mandatory process steps/deliverables via an anonymous survey of our entire new hire population this year.  Now that we have the data, I am not sure how to do the MSA.  I don’t think it makes sense to have the same group retake the survey for repeatability, the answers will be different based on each person’s experience (which rules out reproducibility), and there is not a paper trail to check the accuracy of their responses.
    Any thoughts?



    Its tricky to design MSA for this kind of data, but we still can do this. I presume that the data is confidential, but if you can give me some insight about this data like #of questions, # of people, I will try to design MSA for you. You can mail me at [email protected]


    Kim Niles

    Dear Stacy:
    One of the prerequisites for MSA is having a stable process.  It’s hard to tell from your email but I’d guess you don’t have one.  However, I’d also guess that there are other ways to accomplish what you want to accomplish (assure high confidence that you can hire, train, and manage workers so as not to affect production quality right?). 
    Make sure you’ve characterized the process fully so that anyone can fully understand ownership, SIPOC, flow, step criteria, etc.  Make sure you’ve measured all the key steps against specs or criteria.  Make sure you understand the effect of all key variables that can change (temporary vs. permanent, shift differences, etc??).  Make sure you’ve documented and trained against all the key aspects of the process. 
    I hope that helps.
    KN – 



    There are two analyses that need to be conducted to asses survey results for useability.  The first is reliability (how consistent are my measurements), the second is validity (am I measuring what I think I’m measuring).  You need to do both for an MSA, but this response only addresses the first issue, reliability.
    Typically, an internal reliability analysis is conducted to assess the accuracy (e.g., repeatability, consistency) of surveys, assuming you a measuring a single construct – it removes the need to conduct a re-test as it would yield similar results.  If you are measuring more than one construct, group the appropriate survey items together and run a  separate reliability analysis for each construct.  (Imagine that each survey item is a process, and group similar processes.)   The results range from 0 to 1 (like a correlation coefficient);  above 0.7 is considered good, 0.8+ is very good, and 0.9+ is exceptional.  Squaring the result and converting to a percentage will yield the amount of shared variance among all items, and thus, the degree of accuracy.  The remaining variance is due to sampling error, question design, bias, etc.  Since you are collecting data from all new-hires, sampling error is not an issue.  Therefore, a low reliability estimate would indicate that you aren’t asking the right questions, the questions aren’t well formulated, certain questions are biased, etc.  SPSS, among other software, will calculate reliability (you can also do it manually: not hard, just takes some time).



    You cannot run an MSA on survey data let’s clear that up right now!
    Why would you weant to? The raw data is what it is you cannot go back to these people and re ask the same questions. If you had created the document so that you had contained therein two or three question that expected the same answer then it would have been possible.
    I believe you can run an MSA on the cycle time data

Viewing 5 posts - 1 through 5 (of 5 total)

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