# Regressing a differenced variable on a lagged variable. How can I fix the error in R?

I have a time series (std) of 324 observations with no missing values, starting from January 1987 and ending in December 2013.

I want to regress via OLS the one in the question.

In R, the code:

 lm(dstd ~ lstd)


Where 'dstd' is the differenced variable, via:

 dstd<-diff(std,differences=1)


And lstd is the lagged variable via:

 lstd<-lag(std, k=1)


Does not work, as it gives the following error:

 Error in model.frame.default(formula = dstd ~ lstd, drop.unused.levels = TRUE) :
variable lengths differ (found for 'lstd')


Which is, of course, natural as the length of 'dstd' is 323 while the length of 'lstd' is 324.

Searching for the error, na.omit has been suggested as the answer for other models. However, no combinations of 'na.omit' or 'na.exclude' would work. This video (at the relevant time) says ' you need to change these', referring to these variables.

I have tried excluding while forming another object, as well as subsetting in the regression model, yet none of these seem to work.

What code would I need to fix this?

Here is a Minimal Reproducible Example:

An object 'stdshort' is created with the same characteristics of the parent object, via this code:

 > stdshort<-dput(window(std, 1987, 1989, 12))
structure(c(4.5, 4.7, 4.2, 4.4, 3.9, 3.9, 3.7, 3.7, 3.4, 3.6,
3.5, 3.1, 3.5, 3.3, 3.7, 3.7, 3.7, 3.8, 3.6, 3.5, 3.5, 3.3, 3.5,
3.5, 3.4), .Tsp = c(1987, 1989, 12), class = "ts")


The following code replicates the steps previously mentioned: differencing, lagging and linear regression.

First, a differenced object is created. The object dtstdshort's structure looks like this:

 > dstdshort<-diff(stdshort,differences=1)
> dput(dstdshort)
structure(c(0.2, -0.5, 0.2, -0.5, 0, -0.2, 0, -0.3, 0.2, -0.1,
-0.4, 0.4, -0.2, 0.4, 0, 0, 0.0999999999999996, -0.2, -0.1, 0,
-0.2, 0.2, 0, -0.1), .Tsp = c(1987.08333333333, 1989, 12), class = "ts")


A following object, lstdshort, for 1 period lagging is constructed. This is the structure:

 > lstdshort<-lag(stdshort,k=1)
> dput(lstdshort)
structure(c(4.5, 4.7, 4.2, 4.4, 3.9, 3.9, 3.7, 3.7, 3.4, 3.6,
3.5, 3.1, 3.5, 3.3, 3.7, 3.7, 3.7, 3.8, 3.6, 3.5, 3.5, 3.3, 3.5,
3.5, 3.4), .Tsp = c(1986.91666666667, 1988.91666666667, 12), class = "ts")


The regression of those two objects as explained before, gives the following error (with code):

 > lm(dstdshort ~ lstdshort)
Error in model.frame.default(formula = dstdshort ~ lstdshort, drop.unused.levels = TRUE) :
variable lengths differ (found for 'lstdshort')

• This question appears to be off-topic because it is about programming without a reproducible example. – Glen_b -Reinstate Monica Sep 12 '14 at 11:35
• Would uploading the data make it better? – erasmortg Sep 12 '14 at 11:48
• I voted to leave this question open because although it is phrased in terms of "code" it reflects a very basic problem that I believe will have a universal (code-independent) answer. – whuber Sep 12 '14 at 13:40
• I'd suggest the smallest reasonable example with the same issue. – Glen_b -Reinstate Monica Sep 12 '14 at 21:52
• Added relevant Minimal Reproducible Example as requested. – erasmortg Sep 13 '14 at 8:22

The function ts.intersect sets up the variables for you.
stdshortcomb <-  ts.intersect(dstdshort,lstdshort)