large standard deviation

Six Sigma – iSixSigma Forums Old Forums Healthcare large standard deviation

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
  • #25178


    Got a question from a green belt and need help answering.   We are getting consistent standard deviations larger than the mean from different areas of our facility.  Each measure relates to the same process, just in a different location.  The sigma levels are fair to poor.  Capability index seems to be way too low.  Just wonder what the thought might be.  For example:
    Average processing time = 24 minutes
    Std. Deviation = 34.35 and sigma = 2.16 on a goal of 20 minutes.
    Any help?



    My first thought is that the processes are not the same. I would look at the median, also is there a way to analyze the fastest and slowest times to see what the difference is?
    If you create a histogram by the different areas what does it look like?
    What is the mean / median by each different location? What location has the largest variation?



    I agree with GDS. If the processes are same, in normal course, it should not have large deviations. What is the spread? What are the variable? Would suggest a DoE on the process and measurement capability. How reliable is the measurement? Maybe there are certain factors that creep in at that location. DoE should help, I guess.
    All the best.



    One of the main points of six sigma is to drive down standard deviation.  Since your overall standard deviation is so large, it makes sense that your capability index is very low, being that capability index is inversely related to standard deviation.  In other words, you have a project that needs a lot of attention and standardization, so that you can drive down the standard deviaition.  My guess is that there is much wasted effort and time on the process.  If you need any more ideas, please feel free to email me.



    Is  your process in statistical control? Do you have the SPC chart? Can you see any trends or cycles? 



    Identify from your various groups their mean and SD, and what the spread is for the mean and sd.  Look closely at those that the mean is spot on but the SD is wide or the mean is out but the Sd is tight
    with this information you should be able to identify the next process to take.


    Robert Butler

      As others have said – first plot the data – location-by-location and then with all of the data on the same histogram. Do the same thing again using boxplots – Look at the plots – you are talking about processing time.  This is a variable which probably isn’t normally distributed to begin with.  From the boxplots check to see how well all of the locations line up with one another – are there one or more whose population is shifted relative to the majority?  Are they scattered all over the map? A check of this type will give you an understanding of which groups are behaving in a similar fashion and which ones aren’t.
    In addition to the location by location computation of standard deviations and means compute the medians as well.  How well do the medians and the means agree? If there are large differences across all locations then the standard deviation of the raw data probably isn’t the best way to characterize the system variability. 
      If all groups are not the same don’t waste time on figuring Cpk or standard deviation – your first priority is identification of special causes.  If they are all about the same and the standard deviation is representative then you are going to have to break down the systems and identify major existing sources of variation.

Viewing 7 posts - 1 through 7 (of 7 total)

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