LISTING OF ALL SUBSETS AND SELECTION OF BEST SUBSET IN A FINITE DISTRIBUTED LAG MODEL USING METEOROLOGICAL DATA

This paper develops algorithms for listing and estimation of all subsets in a finite distributed lag

model, with a view to select the best subset from the listed subsets thereby removing the

parameters that contribute insignificantly to the model and more so reduce the problem of multicollinearity.

We

describe

the

method

of

estimation

for

full

and

subset

finite

distributed

lag

models.

In

order

to select the best order for full and subset finite distributed

lag models,

Akaike information

criterion (AIC) is adopted. The estimation technique is illustrated with respect to a meteorological

series where rainfall is the dependent variable and temperature is the explanatory variable. The best

order for the full finite distributed lag model is achieved at lag six (6) and the 2

q

-1 for all subsets 

gave a total of sixty three (63) subsets that is 2

6

-1. The best order having estimated the sixty three 

equations occurred at the 31

st

 subsets. This led to the removal of insignificant parameters in the