Prev: Robustifying batch Excel import
Next: Patient visit follow up
From: Dimitri Shvorob on 15 Oct 2009 07:07
I am PROC IMPORT'ing a large number of common-format Excel spreadsheets and
concatenating resulting datasets. Occasionally, a numeric column in a
spreadsheet has no values (empty cells) and is interpreted by SAS as having
char type, tripping up subsequent data-step concatenation with 'Variable x
has been defined as both character and numeric' error.

PROC APPEND with FORCE option ploughs through, to encounter a different,
familiar problem: if char column s has shorter values in the 'base' dataset
than in the 'data' one, column length of s is determined by 'base', and
strings in 'data' truncated. (The same would happen with 'data ab; set a b;
run;', of course).

I am guessing that

1. I cannot make PROC IMPORT see that a variable is numeric, not char, so
PROC APPEND it is. (With SAS 8, I cannot try the Excel libname route).

2. LENGTH issue can be resolved by 'seeding' the concatenation with an
empty dataset, created by a data step with a custom (alas) series
of 'length' statements.

Can anyone suggest a better way?

Thank you.
 | 
Pages: 1
Prev: Robustifying batch Excel import
Next: Patient visit follow up