DWP Data

Can Southern California's water supply in future years be 
predicted from past data? One factor affecting water 
availability is stream runoff. If runoff could be predicted, 
engineers, planners and policy makers could do their jobs 
more efficiently. Multiple linear regression models have been 
used in this regard.

The following dataset contains 43 years worth of 
precipitation measurements (in inches) taken at six sights in 
the Owens Valley labeled APMAM (Mammoth Lake), APSAB (Lake 
Sabrina), APSLAKE (South Lake), OPBPC (Big Pine Creek), OPRC 
(Rock Creek), and OPSLAKE, and stream runoff volume (measured 
in acre-feet) at a sight near Bishop, California (labeled 
BSAAM).

 Year   APMAM   APSAB  APSLAKE  OPBPC   OPRC   OPSLAKE   BSAAM   
 1948    9.13    3.58    3.91    4.10    7.43    6.47    54235   
 1949    5.28    4.82    5.20    7.55   11.11   10.26    67567   
 1950    4.20    3.77    3.67    9.52   12.20   11.35    66161   
  ...

 1986    4.93    3.26    4.58   26.47   15.33   26.46   118144   
 1987    5.99    2.76    3.98    4.80    6.85    6.36    61229   
 1988    6.83    6.82    5.18    7.20    9.01    9.88    58942   
 1989    8.80    5.06    4.92    8.05    9.60    9.58    53965   
 1990    7.10    5.06    6.05    5.80    6.50    8.41    49774   

EXERCISES

Can precipitation predict runoff volume? There is certainly 
good reason to think so. The major source of runoff is 
precipitation, although there may be some time lag related to 
season. The following hints might prove useful. 1) Try to 
select a set of "important" explanatory variables.  2) There 
are very strong linear relationships between the explanatory 
variables, giving rise to the phenomenon of 
multicollinearity.  3) Some regression models contain 
outliers. Remove them and reexamine your model.  4) Since you 
are dealing with time ordered data check for the presence of 
autocorrelation. 

George Michailides   gmichail@stat.ucla.edu