QQPLOT Statement: CAPABILITY Procedure

Graphical Estimation

You can use Q-Q plots to estimate shape, location, and scale parameters and to estimate percentiles. If you are working with a normal Q-Q plot, you can also estimate certain capability indices.

Shape Parameters

Some distribution options in the QQPLOT statement require that you specify one or two shape parameters in parentheses after the distribution keyword. These are summarized in Table 5.72.

You can visually estimate a shape parameter by specifying a list of values for the shape parameter option. A separate plot is displayed for each value, and you can then select the value that linearizes the point pattern. Alternatively, you can request that the plot be created using an estimated shape parameter. See the entries for the distribution options in the section Dictionary of Options. for details on specification of shape parameters. Example 5.22 and Example 5.23 illustrate shape parameter estimation with lognormal and Weibull Q-Q plots.

Note that for Q-Q plots requested with the WEIBULL2 option, you can estimate the shape parameter c from a linear pattern using the fact that the slope of the pattern is $\frac{1}{c}$. For an illustration, see Example 5.23.

Table 5.72: Shape Parameter Options for the QQPLOT Statement

Distribution Keyword

Mandatory Shape Parameter Option

Range

BETA

ALPHA=$\alpha $, BETA=$\beta $

$\alpha >0$, $\beta >0$

EXPONENTIAL

None

 

GAMMA

ALPHA=$\alpha $

$\alpha >0$

GUMBEL

None

 

LOGNORMAL

SIGMA=$\sigma $

$\sigma >0$

NORMAL

None

 

PARETO

ALPHA=$\alpha $

$\alpha >0$

POWER

ALPHA=$\alpha $

$\alpha >0$

RAYLEIGH

None

 

WEIBULL

C=c

a > 0

WEIBULL2

None

 


Location and Scale Parameters

When the point pattern on a Q-Q plot is linear, its intercept and slope provide estimates of the location and scale parameters. (An exception to this rule is the two-parameter Weibull distribution, for which the intercept and slope are related to the scale and shape parameters.) Table 5.73 shows how the intercept and slope are related to the parameters for each distribution supported by the QQPLOT statement.

Table 5.73: Intercept and Slope of Linear Q-Q Plots

 

Parameters

Linear Pattern

Distribution

Location

Scale

Shape

Intercept

Slope

Beta

$\theta $

$\sigma $

$\alpha $ , $\beta $

$\theta $

$\sigma $

Exponential

$\theta $

$\sigma $

 

$\theta $

$\sigma $

Gamma

$\theta $

$\sigma $

$\alpha $

$\theta $

$\sigma $

Gumbel

$\mu $

$\sigma $

 

$\mu $

$\sigma $

Lognormal

$\theta $

$\zeta $

$\sigma $

$\theta $

$\exp (\zeta )$

Normal

$\mu $

$\sigma $

 

$\mu $

$\sigma $

Generalized Pareto

$\theta $

$\sigma $

$\alpha $

$\theta $

$\sigma $

Power Function

$\theta $

$\sigma $

$\alpha $

$\theta $

$\sigma $

Rayleigh

$\theta $

$\sigma $

 

$\theta $

$\sigma $

Weibull (3-parameter)

$\theta $

$\sigma $

c

$\theta $

$\sigma $

Weibull (2-parameter)

$\theta _0$ (known)

$\sigma $

c

$\log (\sigma )$

$\frac{1}{c}$


You can enhance a Q-Q plot with a diagonal distribution reference line by specifying the parameters that determine the slope and intercept of the line; alternatively, you can request estimates for these parameters. This line is an aid to checking the linearity of the point pattern, and it facilitates parameter estimation. For instance, specifying MU=3 and SIGMA=2 with the NORMAL option requests a line with intercept 3 and slope 2. Specifying SIGMA=1 and C=2 with the WEIBULL2 option requests a line with intercept $\log (1) = 0$ and slope $\frac{1}{2}$.

With the LOGNORMAL and WEIBULL2 options, you can specify the slope directly with the SLOPE= option. That is, for the LOGNORMAL option, specifying THETA=$\theta _0$ and SLOPE=$\exp (\zeta _0)$ gives the same reference line as specifying THETA=$\theta _0$ and ZETA=$\zeta _0$. For the WEIBULL2 option, specifying SIGMA=$\sigma _0$ and SLOPE=$\frac{1}{c_0}$ gives the same reference line as specifying SIGMA=$\sigma _0$ and C=$c_0$.

For an example of parameter estimation using a normal Q-Q plot, see Adding a Distribution Reference Line. Example 5.22 illustrates parameter estimation using a lognormal plot, and Example 5.23 illustrates estimation using two-parameter and three-parameter Weibull plots.

Theoretical Percentiles

There are two ways to estimate percentiles from a Q-Q plot:

  • Specify the PCTLAXIS option, which adds a percentile axis opposite the theoretical quantile axis. The scale for the percentile axis ranges between 0 and 100 with tick marks at percentile values such as 1, 5, 10, 25, 50, 75, 90, 95, and 99. See Figure 5.41 for an example.

  • Specify the PCTLSCALE option, which relabels the horizontal axis tick marks with their percentile equivalents but does not alter their spacing. For example, on a normal Q-Q plot, the tick mark labeled "0" is relabeled as "50" because the 50th percentile corresponds to the zero quantile. See Figure 5.42 for an example.

You can also estimate percentiles using probability plots created with the PROBPLOT statement. See Output 5.20.1 for an example.

Capability Indices

When the point pattern on a normal Q-Q plot is linear, you can estimate the capability indices $CPU$, $CPL$, and $C_{pk}$ from the plot, as explained by Rodriguez (1992). This method exploits the fact that the horizontal axis of a Q-Q plot indicates the distance in standard deviation units (multiple of $\sigma $) between a measurement or specification limit and the process average.

In particular, one-third the standardized distance between an upper specification limit and the mean is the one-sided capability index CPU.

\[ \mi{CPU} = \frac{USL - \mu }{3\sigma } \]

Likewise, one-third the standardized distance between a lower specification limit and the mean is the one-sided capability index CPL.

\[ \mi{CPL} = \frac{\mu - LSL}{3\sigma } \]

Consequently, if you rescale the quantile axis of a normal Q-Q plot by a factor of three, you can read CPU and CPL from the horizontal coordinates of the points at which the upper and lower specification lines intersect the point pattern. Because $C_{pk}$ is defined as the minimum of CPU and CPL, this method also provides a graphical estimate of $C_{pk}$. For an illustration, see Example 5.24.