error in Matchit - caliper matching using mahalanobis distances? - Matchit

7 Apr 2009

Dear colleagues,

I encountered a problem in using matchit; more particular in the
combination of matching using propensity scores including a caliper
score based on the mahalanobis distance. Below you find a fragment of my
R-output:

Problem:
...
  panel <-  read.table("G:/data 25-03-09 na norm
alleen catiwapi.dat", header=T, quote="", dec=".")
...
  dim(panel) [1] 4032   19
...
  tevredenm.out.cal <- matchit(catiwapi~
inkomen+stedelijk+V66+V63+V61+job+livalone
+                   +zorgen+V44+V61*V63+V61*livalone+V66*job+
+                   V61*job+V63*V61,data=panel,distance="mahalanobis", 
+                   mahvars = c("V61", "V66", "zorgen"),
caliper=0.0005)
Error: subscript out of bounds

I found this error, despite many efforts to solve it. Apparantly, R
cannot extract the individual variables from the data matrix (despite
the fact I atta
In the end I found one workaround which didn't produce the error, but I
have no clue why this is. Did anyone else encounter this particular
error?

Solved by:
...
  panels <-  read.table("G:/data 25-03-09 na
norm alleen catiwapi.dat", header=T, quote="", dec=".")
> panel <-  panels[1:4032,1:19]
...
  tevredenm.out.cal <- matchit(catiwapi~
inkomen+stedelijk+V66+V63+V61+job+livalone
+                   +zorgen+V44+V61*V63+V61*livalone+V66*job+
+                   V61*job+V63*V61,data=panel,distance="mahalanobis", 
+                   mahvars = c("V61", "V66", "zorgen"),
caliper=0.0005)
works as I want it

warm greeting from the Netherlands,

Peter Lugtig
Methoden en technieken/ methods and statistics
Utrecht University