Statistics questions

hmm, guess why this is a problem for the search function.. thus without an good abstract in the body and a good subject you won't get many experts reading your question.

Even though your question is incoherent, I can tell you for sure that your model doesn't make any sense.

• On Cake_5, if you skip Step_1 what exactly do you prepare in Step_2?  And how do you melt mix and cream?

• Your steps shouldn't have equal weight.  If you skip Step_2, you will get a perfect cake that won't come out of the pan.

• If you skip Step_6, _7, or _8, the cake should get a score of zero not 7/8.

If you can explain what your are trying to do in words we can help.

d-glitch:

Thank you for chiming in... to keep things simple, I am using the "baking cake" scenario as an analogy.   Walking anyone through the actual scenario would be too confusing (and I probably would violate company policy).   Anyhow, please allow me to provide a little more information.

Based on the analogy, there are 10 activities that are being performed by test participants.   Btw, some of the activities are more important that others.   But for the sake for argument, that is not important as part of this discussion.

So, for baking activity number one, anyone who performs this step receives a check in the box (also, for the point of this discussion, it's not important how well the person did it).   So, it's merely a binary check.   Person performed the activity or or he/she didn't do it.

Upon completion of all 10 steps, I tally up the "did perform" (horizontally) and I can make some form assessment on the actual activity.   That, something was performed 90%.

Now, when looking at the final product (i.e., the cake), I need to look at the matrix and calculate the "checks" vertically.   That is, for cake #1, were any of the steps missed?   Again, I understand that some steps are more important that others (i.e., adding flower/eggs is more important that greasing the pan).    At any rate, assuming equal weight, if I were to make **predictions about future cakes**, can I merely sum up the checks vertically and that look at the individual totals?    This is where my colleague made the point of only being 35% accurate (green textbox in the Excel).

Any additional thoughts are welcomed.

Thank you,
EEH

Ian - thank you for your assistance in this matter.   Your feedback was extremely helpful.

Regards,
EEH

Ian:

Final question... where does the "0.994884" originate from?   Are we referring to Six Sigma probability here?

Thanks,
EEH

Hi there EEH,

Sorry I missed your request.

The 0.994884 has nothing to do with six sigma (which is a normal probability value). In fact it doesn't assume any distribution of values except pass/fail probability of 0.95.  

I assumed you were interested in a combined probability of success of 0.95, though the working below can easily be modified for other combined probabilities and other numbers of sub-tasks making up the total success.  All this assumes is that (as I said before) the sub-tasks are independent.  I would suggest extreme caution here as I have a feeling that the sub-tasks are likely NOT to be independent.


For 10 independent event with equal chance of success (xi), and the probability of total success (xtotal) (that is all 10 successes) is just xi raised to the 10th power.

ptotal = xi ^ 10

log(ptotal) = 10 log(xi)
log(xi) = log(ptotal) / 10
xi = antilog( log(ptotal) / 10)

xi = antilog( log(0.95) / 10)
xi = antilog( -0.0222764 / 10)
xi = antilog( -0.00222764)
xi = 0.994884

Select allOpen in new window

Ian

Reasonable attempt to help with roughly defined problem.