Generate lags R

I hope this is basic; just need a nudge in the right direction.

I have read in a database table from MS Access into a data frame using RODBC. Here is a basic structure of what I read in:

PRODID PROD Year Week QTY SALES INVOICE

Here is the structure:

str(data)
'data.frame': 8270 obs. of  7 variables:
 $ PRODID  : int  20001 20001 20001 100001 100001 100001 100001 100001 100001 100001 ...
 $ PROD    : Factor w/ 1239 levels "1% 20qt Box",..: 335 335 335 128 128 128 128 128 128 128 ...
 $ Year    : int  2010 2010 2010 2009 2009 2009 2009 2009 2009 2010 ...
 $ Week    : int  12 18 19 14 15 16 17 18 19 9 ...
 $ QTY     : num  1 1 0 135 300 270 300 270 315 315 ...
 $ SALES   :开发者_高级运维 num  15.5 0 -13.9 243 540 ...
 $ INVOICES: num  1 1 2 5 11 11 10 11 11 12 ... 

Here are the top few rows:

head(data, n=10)
   PRODID           PROD Year Week QTY  SALES INVOICES
1   20001      Dolie 12" 2010   12   1  15.46        1
2   20001      Dolie 12" 2010   18   1   0.00        1
3   20001      Dolie 12" 2010   19   0 -13.88        2
4  100001 Cage Free Eggs 2009   14 135 243.00        5
5  100001 Cage Free Eggs 2009   15 300 540.00       11
6  100001 Cage Free Eggs 2009   16 270 486.00       11
7  100001 Cage Free Eggs 2009   17 300 540.00       10
8  100001 Cage Free Eggs 2009   18 270 486.00       11
9  100001 Cage Free Eggs 2009   19 315 567.00       11
10 100001 Cage Free Eggs 2010    9 315 569.25       12 

I simply want to generate lags for QTY, SALES, INVOICE for each product but I am not sure where to start. I know R is great with Time Series, but I am not sure where to start.

I have two questions:

1. I have the raw invoice data but have aggregated it for reporting purposes. Would it be easier if I didn't aggregate the data?

2. Regardless of aggregation or not, what functions will I need to loop over each product and generate the lags as I need them?

In short, I want to loop over a set of records, calculate lags for a product (if possible), append the lags (as they apply) to the current record for each product, and write the results back to a table in my database for my reporting software to use.

There is most likely a more elegant way to do this.

First read in the data:

h <- 'row   PRODID           PROD Year Week QTY  SALES INVOICES
1   20001      Dolie12 2010   12   1  15.46        1
2   20001      Dolie12 2010   18   1   0.00        1
3   20001      Dolie12 2010   19   0 -13.88        2
4  100001 CageFreeEggs 2009   14 135 243.00        5
5  100001 CageFreeEggs 2009   15 300 540.00       11
6  100001 CageFreeEggs 2009   16 270 486.00       11
7  100001 CageFreeEggs 2009   17 300 540.00       10
8  100001 CageFreeEggs 2009   18 270 486.00       11
9  100001 CageFreeEggs 2009   19 315 567.00       11
10 100001 CageFreeEggs 2010    9 315 569.25       12'
dat <- read.table(textConnection(h),T)

next we calculate the lagged variables. This line of code splits up the data by PROD, then puts NAs as the top row, and drops the last row, resulting in a lag of 1.

new_vars <-do.call(rbind,rev(by(dat[,c(6,7,8)],dat$PROD,function(x) rbind(NA,x[-nrow(x),]))))

Show them in the console:

> cbind(dat,new_vars)
               row PRODID         PROD Year Week QTY  SALES INVOICES QTY  SALES INVOICES
Dolie12.1        1  20001      Dolie12 2010   12   1  15.46        1  NA     NA       NA
Dolie12.2        2  20001      Dolie12 2010   18   1   0.00        1   1  15.46        1
Dolie12.3        3  20001      Dolie12 2010   19   0 -13.88        2   1   0.00        1
CageFreeEggs.1   4 100001 CageFreeEggs 2009   14 135 243.00        5  NA     NA       NA
CageFreeEggs.4   5 100001 CageFreeEggs 2009   15 300 540.00       11 135 243.00        5
CageFreeEggs.5   6 100001 CageFreeEggs 2009   16 270 486.00       11 300 540.00       11
CageFreeEggs.6   7 100001 CageFreeEggs 2009   17 300 540.00       10 270 486.00       11
CageFreeEggs.7   8 100001 CageFreeEggs 2009   18 270 486.00       11 300 540.00       10
CageFreeEggs.8   9 100001 CageFreeEggs 2009   19 315 567.00       11 270 486.00       11
CageFreeEggs.9  10 100001 CageFreeEggs 2010    9 315 569.25       12 315 567.00       11

You can use back() function from the PERregress library on a single variable. You can also create multiple lags. I typically use this function for auto-correlation in regression analysis.