R help - May 2011 - using predict.lm function

Dear all,

I'm fitting a linear model with numerous lag terms of the response variable
[i.e. y(t-1), y(t-2),y(t-3)...,] and other explanatory variables [x(1), x(2),
x(3),....]- which go into my design matrix X.

I'm fitting the linear model: lm(Y ~ X, ...).

I would like to use the predict.lm function however the future predictions of Y
are dependent upon previous predictions of Y [i.e. the response lag terms]. Does
anyone know how I would go about using predict.lm to make future forecasts of Y?
Or, does this have to be writting as a for loop that recursively updates the lag
terms for each future prediction of y(i)? The data is a time series and the
model is only calculated once in order to make the future predictions.

Thanks,
Dave
	[[alternative HTML version deleted]]

On May 22, 2011, at 12:35 PM, Dave Evens wrote:
>
> Dear all,
>
> I'm fitting a linear model with numerous lag terms of the response  
> variable [i.e. y(t-1), y(t-2),y(t-3)...,] and other explanatory  
> variables [x(1), x(2), x(3),....]- which go into my design matrix X.
>
> I'm fitting the linear model: lm(Y ~ X, ...).
>
> I would like to use the predict.lm function however the future  
> predictions of Y are dependent upon previous predictions of Y [i.e.  
> the response lag terms]. Does anyone know how I would go about using  
> predict.lm to make future forecasts of Y? Or, does this have to be  
> writting as a for loop that recursively updates the lag terms for  
> each future prediction of y(i)? The data is a time series and the  
> model is only calculated once in order to make the future predictions.
There is a 'forecast' package.

-- 

David Winsemius, MD
West Hartford, CT