![]() | This article is rated Start-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | ||||||||||||||||||||
|
The p-chart doesn't exist
edit>In industrial statistics, the p-chart is a type of control chart that is very similar to the X-bar chart >except that the statistic being plotted is the sample proportion rather than the sample mean.
The p-chart doesn't exist: it is also an X-bar chart.
X can be Gauss or Bernouilli distributed.
Martin Segers
- Im a sorry, you are mistaken. The p-chart for the proportion of nonconforming units, based on the binomial distribution, is a standard process control chart. Here for example is an ASQ article on it. Numerous SPC guides refer to it as such. Just off the top of google books, from the Quality Technician's Handbook, "The most commonly used attribute control chart is the p chart". The article could do with some references though. Cheers, Jayen466 13:31, 28 October 2008 (UTC)
- Gauss process:
- The mean value of the X-bar distribution is µ.
- We speak here of an X-bar chart and not of a µ-chart.
- Bernouilli process (X = 0 or 1):
- The mean value of the X-bar distribution is p.
- Also here we have to speak of an X-bar chart and not of a p-chart.
- That this chart is called "p-chart" is due to historical circumstances but is misleading.
- Martin Segers
- p, in the sense used here, is not defined as an average. p stands for the proportion of non-conforming items in a sample. To give a simplified example, a process may produce 100 units each day. Each day, the 100 parts produced are inspected, and the proportion p of nonconforming units is noted and plotted on the chart. For a given (constant) probability of producing a non-conforming unit, the statistic p follows a binomial distribution. The standard deviation of this binomial distribution is used to calculate the control limits, to distinguish between random fluctuations of p (underlying probability of a unit not conforming remains constant), and special-cause variation in p (p values in the extreme tails of the binomial distribution are assumed to reflect special-cause variation, i.e. it is assumed that the probability of the process producing a nonconforming unit has increased or decreased). Of course you could say that p is an average of 0 and 1 values assigned to the individual process outcomes, but overwhelming practice simply defines p as the proportion of non-conforming units in a sample, and as a statistic in its own right that follows the binomial distribution. I believe in Europe, people do tend to use the x designation for this chart; but in the US and the UK, the p designation is standard (the four basic types of attribute chart being p, np, c and u). Cheers, Jayen466 13:26, 1 November 2008 (UTC)
n = sample size
variable p : mean = pbar, st.dev. = square(pbar(1-pbar)/n)
variable np : mean = npbar, st.dev. = square(npbar(1-pbar))
mother population : mean = pbar, st.dev. = square(pbar(1-pbar))
Which symbol is used for the mother population variable?
Martin Segers, 30 Nov 2008
>p, in the sense used here, is not defined as an average. p stands for the proportion of non-conforming >items in a sample.
The mean value of 0 and 1 is 0.5.
The proportion of 1's is also 0.5.
So if the numbers are or 0 or 1 then the proportion 1's and the mean are the same.
Martin Segers, 1 Dec 2008
- I understand this comment, but there are 265 books in google books referring to the p-chart in connection with SPC. What to do? We cannot improve practice, we are supposed to mirror it. Regards, Jayen466 21:43, 1 December 2008 (UTC)