iSixSigma

MSA with large databases

Six Sigma – iSixSigma Forums Old Forums General MSA with large databases

Viewing 3 posts - 1 through 3 (of 3 total)
  • Author
    Posts
  • #51008

    newbie
    Participant

    What would be the approach for the MSA when the Y data is coming from large data warehouses?  Thanks!

    0
    #176140

    Clive
    Participant

    For database data I’ve always worked on if you can prove that the data enetered into the database is correct and that your extraction methods get the data correctly then that’s your MSA.
    On Proof of extraction that’s basic software testing. On the validity of values in the system  that will also be software testing methods as well on the data entry and you would also have to decide if you require to do an MSA on the entry/data collection method. That will depend on the type of data in the database.

    0
    #176164

    Remi
    Participant

    Hai newbie,
    DMAIC3 is intended to make sure that when later on in your project you analyse the Y-data this analysis tells you something about the process you want/need to improve. So DMAIC3 could be seen as: how much are you sure that the Y-value you get is ‘correct’ and not influenced by other things (like the measuring action).
    If you get data from a data warehouse:

    the data has to be measured before it is recorded in the warehouse => gage r&R
    the data is transported from the measurement system into the warehouse: check this for correctness
    the data is kept in the warehouse: check this
    the data is send to somebody who asks for it (=you): check this
    Other issues:

    data warehouses (old ones) don’t have infinite storage capacity: how do they overwrite old data => check this
    when putting data in/out: what happens at beginning and tails of the dataset (often I encountered some ‘strange’ data there)
    did somebody already include summary data in your data set (I encountered data where an extra row of shift-average was inserted in the data after each shift)
    how many missing data is there in the data set. If much than doubt the usefullness
    are the data formats defined correctly
    can you assume that the persons who have to ensure that you get the correct data did their work perfectly correct
    Each check you do not do is a risk you take; each check you do is seen as ‘wasted effort’ if you find out it is ok. But that’s life.
    Hope this helps

    0
Viewing 3 posts - 1 through 3 (of 3 total)

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