*Missing pairwise - http://stats.stackexchange.com/q/105006/1036.

DATASET CLOSE ALL.
OUTPUT CLOSE ALL.

*SPSS Code.
DATA LIST FREE /A B C.
BEGIN DATA
1  1  .
2  .  2
.  3  3
4  4  4
5  5  5
END DATA.

CORRELATIONS A B
  /MISSING PAIRWISE
  /STATISTICS XPROD.

REGRESSION VARIABLES = A,B
  /DEPENDENT = A
  /METHOD = ENTER B
  /MISSING PAIRWISE.

REGRESSION VARIABLES = A,B
  /DEPENDENT = A
  /METHOD = ENTER B
  /MISSING LISTWISE.

*R code.
BEGIN PROGRAM R.
a <- c(1,2,NA,4,5)
b <- c(1,NA,3,4,5)

#This is how I thought SPSS would do it.
covP <- cov(a,b,use="pairwise.complete.obs")
beta <- covP / var(b, na.rm=TRUE)
int  <- mean(a, na.rm=TRUE) - beta*(mean(b, na.rm=TRUE))
beta
int
END PROGRAM.
*These arent the same, not sure how SPSS does it.