To ensure that your results are valid, consider the following guidelines when you collect data, perform the analysis, and interpret your results.
- The data should be continuous
Continuous data are measurements that may potentially take on any numeric value within a range of values along a continuous scale, including fractional or decimal values. Common examples include measurements such as length, weight, and temperature.
If you have attribute data, such as counts of defectives or defects, use Binomial Capability
Analysis or Poisson Capability
- Collect enough data to obtain reliable estimates of process capability
- Try to collect at least 100 total data points (subgroup size*number of subgroups), such as 25 subgroups of size 4, or 35 subgroups of size 3. If you do not collect a sufficient amount of data over a long enough period of time, the data may not accurately represent different sources of process variation and the estimates may not indicate the true capability of your process.
- Data should be collected in rational subgroups, if possible
- A rational subgroup is a small sample of similar items that are produced over a short period of time and that are representative of the process. Observations for each subgroup should be collected under the same inputs and conditions, such as personnel, environment, or equipment. If you do not collect rational subgroups, the variation in the subgroups may reflect special causes rather than the natural, inherent variation of the process.
- The process must be stable and in control
- If the current process is not stable, then the capability indices cannot be reliably used to assess the future, ongoing capability of the process. If you are unsure whether your process is in control, use Xbar-S Chart or Normal Capability
Sixpack to evaluate process stability before you perform this analysis.
- The data should follow a normal distribution
- The process capability estimates for this analysis are based on the normal distribution. If the data are not normally distributed, the capability estimates will not be accurate for your process. If your data are nonnormal, you can transform them using the Box-Cox transformation or the Johnson transformation, which are included in the Transform options of this analysis. To determine whether your data are normal, or whether a transformation will be effective for nonnormal data, use Individual
Distribution Identification. If your data are nonnormal and a transformation is not effective, consider using Nonnormal