Step 3: Displaying the Trend Chart

Graphical Enhancements

Step 3: Displaying the Trend Chart

The third step is to create a trend chart with the SHEWHART procedure, as follows:

   title 'Trend Chart for Diameter';
   symbol v=none c=salmon w=5;
   proc shewhart history=regdata;
      xchart diameter*hour /
         stddeviations
         trendvar = fitted
         cneedles = black
         wtrend   = 1
         split    = '/'
         nolegend
         cframe   = vibg
         cinfill  = ywh;
      label diameterx = 'Residual Mean/Fitted Mean';
   run;

The chart is shown in Figure 47.26. The values of FITTED are plotted in the lower half of the trend chart. The upper half of the trend chart is an $\bar{X}$ chart for the residual means (DIAMTERX - FITTED). The CNEEDLES= option specifies that the residuals are to be represented by vertical bars as deviations from the central line. The $\bar{X}$ chart in Figure 47.26 shows that, after accounting for the trend, the mean level of the process is in control.

Figure 47.26: Trend Chart for Diameter Data

If the data are correlated in time, you can use the ARIMA or AUTOREG procedures in place of the REG procedure to remove autocorrelation structure and display a control chart for the residuals; for an example, see "Autocorrelation in Process Data" . Another application of the TRENDVAR= option is the display of nominal values in control charts for short runs; see "Short Run Process Control" .

Chapter Contents
Previous
Next
Top