iSixSigma

What Stat Test?

Six Sigma – iSixSigma Forums Old Forums Software/IT What Stat Test?

Viewing 2 posts - 1 through 2 (of 2 total)
  • Author
    Posts
  • #26229

    boettler
    Member

    I want to create model to predict the number of test conditions (Y) using the Budgeted Cost at Completion (In hours), Project Length (months) and Number of Requirements.  Any suggestion?   Is some of the data ordinal or can I assume the data to be continuous?

    0
    #63754

    McD
    Participant

    Rob
    You sort of have a problem.  Those 3 “independent” variables are essentially all the same thing.  All three are measures of project size, and probably much more highly correlated with each other than with the number of test conditions.
    To be useful, I suspect you need to estimate the number of test conditions as early as possible.  Since cost at completion and project length aren’t known until the end, that leaves you with the number of requirements.  Although the number of requirements typically changes over the life of a project, a lot of organizations labor under the illusion that they can “freeze” requirements early in the process, so perhaps number of requirements, or better, number of requirements at some stage in the process, might be better yet.
    Many organizations use the number of requirements to estimate the cost, some quite successfully.  If you have a reasonably managed (and large enough) organization, project length is simply cost divided by how many people you want to put on it (obviously, within some limits).  So those two aren’t really very good measures, anyway.
    All that said, if you are constantly working at improving your estimating model, you probably have better data on cost than anything else, so cost might be a better substitute.  Of course, you must recognize that the cost doesn’t “cause” the tests.  It really is the requirements. 
    –McD
     

    0
Viewing 2 posts - 1 through 2 (of 2 total)

The forum ‘Software/IT’ is closed to new topics and replies.