Understanding the Probability Law for P Chart Control Limits

Control charts are essential tools in Statistical Process Control (SPC) used to monitor a process over time and identify variation. The P chart is a specific type of control chart used to monitor the proportion of nonconforming or defective items in a sample.

What is a P Chart?

A P chart tracks the proportion (p) of items in a sample that have a specific attribute, typically a defect or nonconformity. For example, it can be used to monitor the proportion of broken bottles in a production line, or the proportion of incorrect invoices in an office process.

Probability Distribution for Proportion Defective

When we take a sample of items and count how many are defective, we are dealing with a scenario where each item can either be defective or not defective. This is a binary outcome. If we assume that each item's defect status is independent of others and the probability of an item being defective is constant for each item in the sample, this fits the characteristics of a Binomial distribution.

Let's look at the characteristics:

• There are a fixed number of trials (the sample size, denoted as 'n').
• Each trial has only two possible outcomes: success (e.g., item is defective) or failure (e.g., item is not defective).
• The probability of success (p, the proportion defective in the population or process) is the same for each trial.
• The trials are independent.

These are exactly the conditions for a Binomial distribution. The number of defective items in a sample of size 'n', where the probability of a single item being defective is 'p', follows a Binomial distribution, denoted as B(n, p).

The proportion of defective items in the sample is calculated as $\hat{p} = \frac{X}{n}$, where X is the number of defective items in the sample and n is the sample size. Since X follows a Binomial distribution, the distribution of $\hat{p}$ is directly derived from the Binomial distribution of X.

Calculating Control Limits for a P Chart

The control limits for a P chart are typically calculated based on the mean and standard deviation of the sample proportion $\hat{p}$.

For a Binomial distribution B(n, p), the mean number of defectives is $np$, and the variance is $np(1-p)$.

The mean proportion defective $\hat{p}$ is estimated by the overall average proportion $\bar{p}$ from historical data.

The standard deviation of the sample proportion $\hat{p}$ is given by $\sigma_{\hat{p}} = \sqrt{\frac{p(1-p)}{n}}$. Using the estimate $\bar{p}$ for p, the estimated standard deviation is $\sqrt{\frac{\bar{p}(1-\bar{p})}{n}}$.

The control limits are usually set at three standard deviations from the mean proportion:

• Upper Control Limit (UCL) = $\bar{p} + 3\sqrt{\frac{\bar{p}(1-\bar{p})}{n}}$
• Lower Control Limit (LCL) = $\bar{p} - 3\sqrt{\frac{\bar{p}(1-\bar{p})}{n}}$

These formulas directly use the properties of the Binomial distribution to establish the expected variation in the sample proportion, allowing for the calculation of appropriate control limits.

While the distribution of $\hat{p}$ approaches a Normal distribution for large sample sizes (due to the Central Limit Theorem), the fundamental probability law governing the number of defectives (and thus the proportion) in a fixed-size sample is the Binomial distribution. Therefore, the control limits for a P chart are derived directly from the Binomial probability law.

Why Not Other Distributions?

• Poisson: Used for counting the number of events in a fixed interval of time or space, suitable for attributes charts like the C chart (number of defects) or U chart (number of defects per unit) when the opportunity for defects is large but the probability of any single defect is small. Not directly applicable to the proportion defective in a sample of fixed size 'n'.
• Normal: While the Binomial distribution can be approximated by the Normal distribution for large samples, the underlying and exact probability law for the count or proportion in a fixed number of binary trials is Binomial.
• Exponential: Used to model the time until the next event in a Poisson process, typically related to reliability or waiting times. Not relevant for counting discrete defective items in a sample.

Conclusion

Based on the nature of the data collected for a P chart (counting defective items in a fixed-size sample, which are binary outcomes), the underlying probability distribution that governs this count and subsequently the proportion is the Binomial distribution. The control limits are derived from the mean and standard deviation properties of this distribution.