Fuzzy*matching numeric variables

[Angeliki Skoura]

Dear Statalisters,


I am struggling with the implementation of fuzzy matching with numerical variables for my research, using the -rangejoin- command of Robert Picard, Roberto Ferrer and Nick Cox's program (rangejoin sales -1000 1000 1000 using "C:\Users\skour\sour\OneDrive\Computer\skoura research\Diff Databases\dataset 1. dta") in order to do the matching with some diviation between the values, but my output shows this error: "op. sys. refuses to provide memory" regardless of the range I set.
What could be my mistake here?

Thank you very much in advance. Please let me know if I need to clarify any further.


Best wishes,
Angeliki

[Daniel Feenberg]

Of course you might actually be running out of memory. Do you get the same message if you cut the dataset down to a few observations? However, I have seen that message when I did not have write permission in the current directory. There are probably other possibilities.

[Angeliki Skoura]

I'll try to cut the dataset down.

Thank you nery much!
Angeliki

[Angeliki Skoura]

I'm sorry to bother you again, but I cut my sample and again it shows me the same error. What could be happening?

Any help or suggestion would be greatly appreciated.

Thank you in advance.
Angeliki

[Daniel Feenberg]

At this point you could consider appealing to Stata tech support. I have some advice general about getting the best response to that.

1) You can't send support your entire program, you must create the smallest program and data possible that demonstrates the problem. 5 lines of code, 2 variables and 3 observations are plenty for most error reports. It takes time to develope a compact demonstration, but otherwise support won't be convinced there is anything but a "probable user error" and you won't get real attention. You will need to send the .do and the resulting log file so they can reproduce the problem using whatever debugging tools they have. Also, while developing the report you are likely to find exactly which option triggers the problem and may discover a work-around or even that it is a user error. That is usually what happens to me.

2) You can cut out one email iteration by adding the commands:

 query compilenumber
  about

to your test program. Be sure to include you license information with the report.

[Angeliki Skoura]

Thank you very much for the immediate response. I'll do as you say.

Best wishes,
Angeliki