Have a look at page 310 of "Advanced Topics in SPC" for comparative ARL
plots. CUSUM does not show any significant improvement over Shewhart
Charts. Wheeler suggests that CUSUM is useful only for maintaining status
quo, not improvement.
"partha" <partho.ss@[EMAIL PROTECTED]
> wrote in message
news:1183453288.856775.216000@[EMAIL PROTECTED]
> On Jul 1, 11:02 am, "Tony" <a_bu...@[EMAIL PROTECTED]
> wrote:
>> It's not hard to see why Wheeler says that CUSUM is a chart that only
its
>> mother would love ...
>>
>> Rather than jumping in blindly, it would be insightful to ask why your
>> client wants to use these charts rather than Shewhart Charts.
>>
>> There is no need to "normalise the variance". Wheeler discusses this
in
>> great detail in the wonderful little book he has written on this topic
>> "Normaility and the Process Behaviour Chart", SPC Press.
>>
>> Dr Tony Burns
>>
> Thanks for your reply. However, I am afraid I can't agree about CUSUM
> charts being totally useless, as you imply. They were invented to
> detect small/gradual changes in the process mean, and both my
> (limited) experience, and that re****ted by others, shows that they do
> achieve this objective quite well, in particular, better/faster/more
> reliably than Shewhart Charts. My question was about the efforts made
> to extend the CUSUM idea to monitor changes in process variance.
>
|
