See MACMA2 in the SAS/QC Sample LibraryBy default, the MACHART statement estimates the process mean () and standard deviation (
) from the data. This is illustrated in the “Getting Started” section of this chapter. However, there are applications in which standard values (
and
) are available based, for instance, on previous experience or extensive sampling. You can specify these values with the MU0=
and SIGMA0= options.
For example, suppose it is known that the metal clip manufacturing process (introduced in Creating Moving Average Charts from Raw Data) has a mean of 15 and standard deviation of 0.2. The following statements specify these standard values:
ods graphics off; symbol h = .8; title 'Specifying Standard Process Mean and Standard Deviation'; proc macontrol data=Clips1; machart Gap*Day / mu0 = 15 sigma0 = 0.2 span = 4 xsymbol = mu0; run;
The XSYMBOL= option specifies the label for the central line. The resulting chart is shown in Output 9.6.1.
Output 9.6.1: Specifying Standard Values with MU0= and SIGMA0=
The central line and control limits are determined using and
(see the equations in Table 9.9). Output 9.6.1 indicates that the process is out-of-control since the moving averages for
Day
=17, Day
=19, and Day
=20 lie below the lower control limit.
You can also specify and
with the variables
_MEAN_
and _STDDEV_
in a LIMITS= data set, as illustrated by the following statements:
data cliplim; length _var_ _subgrp_ _type_ $8; _var_ = 'Gap'; _subgrp_ = 'Day'; _type_ = 'STANDARD'; _limitn_ = 5; _mean_ = 15; _stddev_ = 0.2; _span_ = 4; run; proc macontrol data=Clips1 limits=Cliplim; machart Gap*Day / xsymbol=mu0; run;
The variable _SPAN_
is required, and its value provides the number of terms in the moving average. The variables _VAR_
and _SUBGRP_
are also required, and their values must match the process and subgroup-variable, respectively, specified in the MACHART statement. The bookkeeping variable _TYPE_
is not required, but it is recommended to indicate that the variables _MEAN_
and _STDDEV_
provide standard values rather than estimated values.
The resulting chart (not shown here) is identical to the one shown in Output 9.6.1.