P
pagnonig
Hello,
I've recently read Dr. Wheeler excellent book "Understanding Variation".
In the book he widely uses the I-MR chart because, if my understanding is correct, despite the data distribution it works pretty well to separate noise from signals (I think this comes from the Tchebychev Inequality).
Of course this easy approach to data analysis makes SPC teaching and implementation very fast and straightforward.
But I have one concern. In the book he avoids using a p-chart in some situations where I'd surely use it.
For example: Number of on time deliveries/total deliveries per month
In my opinion this data is perfect for a p-chart with variable sample size (total deliveries per month).
But he says "The probability of a shipment being on time is not constant for all the shipments in any given month"so, the binomial probability model is inappropriate.
Well, I would chart this data with the same approach as if they were Number of NC items/total items per month, using a p-chart.
What do you think about Dr. Wheeler consideration about the binomial model?
To what extend do you agree with him?
Following his advice, would it make sense to chart the number of NC items/total items per month with a I-MR (even with a different number of total items per each month)?
Thank you for your views on this.
Giusepppe
I've recently read Dr. Wheeler excellent book "Understanding Variation".
In the book he widely uses the I-MR chart because, if my understanding is correct, despite the data distribution it works pretty well to separate noise from signals (I think this comes from the Tchebychev Inequality).
Of course this easy approach to data analysis makes SPC teaching and implementation very fast and straightforward.
But I have one concern. In the book he avoids using a p-chart in some situations where I'd surely use it.
For example: Number of on time deliveries/total deliveries per month
In my opinion this data is perfect for a p-chart with variable sample size (total deliveries per month).
But he says "The probability of a shipment being on time is not constant for all the shipments in any given month"so, the binomial probability model is inappropriate.
Well, I would chart this data with the same approach as if they were Number of NC items/total items per month, using a p-chart.
What do you think about Dr. Wheeler consideration about the binomial model?
To what extend do you agree with him?
Following his advice, would it make sense to chart the number of NC items/total items per month with a I-MR (even with a different number of total items per each month)?
Thank you for your views on this.
Giusepppe