Probabilistic record linkage

[MJose Sanchez]

I am trying to run a probabilistic record linkage between two datasets with no common record identifier. I am using the command reclink2. The command is the following:

reclink2 Firmname add1 pobox unit bldg floor using Amadeus_TOTAL, idmaster( stn_AccountHolderName ) idusing( stn_Nombreempresa ) gen( scorematching )

However, the following error message appears:

Going through 7543 observation to assess fuzzy matches, each .=5% complete
2VALORISEAMEL invalid name
post: above message corresponds to expression 1, variable stn_AccountHolderName
r(198);

The value "2VALORISEAMEL" corresponds to one of the observations of the variable stn_AccountHolderName. But the variable stn_AccountHolderName is a string variable (as required by treclink2):

d stn_AccountHolderName

Variable Storage Display Value
name type format label Variable label
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
stn_AccountHo~e str170 %170s official name

I do not know how to solve this. Thank you very much for your help.

[Andrew Musau]

reclink2 is from the Stata Journal (FAQ Advice #12). I do not use this command, but the error message suggests that the observations need to be valid Stata names. See

help naming_conventions

You could bypass this by passing the variable through -strtoname()-, but length limits may still bite you.

replace stn_AccountHolderName= strtoname(stn_AccountHolderName)

See

help strtoname()