The MVPMODEL Procedure

This example illustrates the basic features of the MVPMODEL procedure by using airline flight delay data available from the U.S. Bureau of Transportation Statistics at http://www.transtats.bts.gov. The example applies multivariate process monitoring to flight delays.

Suppose you want to use a principal component model to create and SPE charts to monitor the variation in flight delays. These charts are appropriate because the data are multivariate and correlated.

The following statements create a SAS data set named MWflightDelays to contain the average flight delays for flights that originate in the midwestern United States by airline. The data set contains variables for nine airlines: AA (American Airlines), CO (Continental Airlines), DL (Delta Airlines), F9 (Frontier Airlines), FL (AirTran Airways), NW (Northwest Airlines), UA (United Airlines), US (US Airways), and WN (Southwest Airlines).

data MWflightDelays;
   format flightDate MMDDYY8.;
   label flightDate='Date';
   input flightDate :MMDDYY8. AA CO DL F9 FL NW UA US WN; 
   datalines; 
02/01/07 14.9  7.1  7.9  8.5 14.8  4.5  5.1 13.4  5.1
02/02/07 14.3  9.6 14.1  6.2 12.8  6.0  3.9 15.3 11.4
02/03/07 23.0  6.1  1.7  0.9 11.9 15.2  9.5 18.4  7.6
02/04/07  6.5  6.3  3.9 -0.2  8.4 18.8  6.2  8.8  8.0
02/05/07 12.0 14.1  3.3 -1.3 10.0 13.1 22.8 16.5 11.5
02/06/07 31.9  8.6  4.9  2.0 11.9 21.9 29.0 15.5 15.2

   ... more lines ...   

02/16/07 31.2 20.8 15.2 20.1  9.1 12.9 22.9 36.4 16.4
; 

The observations for a given date are the average flight delays in minutes of flights that depart from the Midwest. For example, on February 2, 2007, F9 (Frontier Airlines) flights departed an average of 6.2 minutes late.