choosing the right model


Does it makes sense to test H_knot : Beta1 = 0 after having rejected H_knot : Beta2 = 0.

So in other words, if it appears that a 2nd order term involving X is useful in fitting a model to the data, does it make sense to test for a simpler model?




TS Contributor
what is the second term, b1=2 and b2=3 for instance and you rejected b2=0? These two parameters can be tested differently with the t-test if that is what you mean.
I mean in a hypothesis test to check to see if a model included b2 is useful.

so it would be

h0: b2 = 0
hA: b2 doesn't equal 0

we reject h0, so it appears that b2 is useful in describing the relationship between x and y.

So know that the model y=b0+b1x1+b2x2 IS indeed useful, does it make sense to check the usefullness of the model y=b0+b1x1?