R中的时间序列回归效率：如何更好地做到这一点？

Question

I am working on time series, and want to check all the lagged differences for significance(and essentially doing a dickey-fuller test by hand) but that's not important. I can do it, but it's really mechanical, and there must be a way to do this more elegantly. Or at least more efficiently. Any ideas?

y <- log.real.gdp.ts

delta.y.t <- diff(y,differences=1)
lag.y <- lag(y, -1)
L1Dy <- lag(delta.y.t, k=-1)
L2Dy <- lag(delta.y.t, k=-2)    
L3Dy <- lag(delta.y.t, k=-3)    
L4Dy <- lag(delta.y.t, k=-4)    
L5Dy <- lag(delta.y.t, k=-5)    
L6Dy <- lag(delta.y.t, k=-6)    
L7Dy <- lag(delta.y.t, k=-7)    
L8Dy <- lag(delta.y.t, k=-8)    
L9Dy <- lag(delta.y.t, k=-9)    
L10Dy <- lag(delta.y.t, k=-10)  
L11Dy <- lag(delta.y.t, k=-11)  
L12Dy <- lag(delta.y.t, k=-12)  

d = ts.union(delta.y.t, lag.y, L1Dy, L2Dy, L3Dy, L4Dy, L5Dy, L6Dy, L7Dy, L8Dy, L9Dy, L10Dy, L11Dy, L12Dy)               ## takes care of NA's

lm.model.III <- lm(delta.y.t~ lag.y + time(lag.y) + L1Dy + L2Dy + L3Dy + L4Dy + L5Dy + L6Dy + L7Dy + L8Dy + L9Dy + L10Dy + L11Dy + L12Dy, data=d)

I'd really like some kind of loop where I can generate 1:n lagged differences, and then some way to insert all n into my linear model, like

 lm.model.III <- lm(delta.y.t ~ lag.y + time(lag.y) + lagged.diffs.mts)

Answer 1

how about

require(zoo)

delta.y.t <- diff(y,differences=1)
lag.y <- lag(y, -1)
L1Dy <- lag(delta.y.t, -(0:12), na.pad=T)

#for any regression you can then access the number of lags you want:
# 0 lag and na.pad=T are crucial

lm(lag.y ~ L1Dy[,1:5])

Hope this helps

-Chris

Answer 2

The package dynlm adds handling of time series operators to R formulas:

The interface and internals of dynlm are very similar to lm , but currently dynlm offers three advantages over the direct use of lm : 1. extended formula processing, 2. preservation of time series attributes, 3. instrumental variables regression (via two-stage least squares). For specifying the formula of the model to be fitted, there are additional functions available which allow for convenient specification of dynamics (via d() and L() ) or linear/cyclical patterns (via trend() , season() , and harmon() ). All new formula functions require that their arguments are time series objects (ie, " ts " or " zoo ").

Here is an example:

library(foreign)
library(zoo)
library(dynlm)

dfKlein = read.dta('http://www.stata-press.com/data/r12/klein.dta')
summary(dfKlein)

zooKlein = as.zoo(dfKlein, order.by = dfKlein$year)
lmKlein = dynlm(consump ~ L(profits, 1) + profits + wagetot, 
                data = zooKlein)
summary(lmKlein)

Note, in particular, that it allows you to specify a vector of lags in the formula object, such as y ~ L(y, 1:4) .