Most changes in new versions add new features and fix bugs rather than
breaking old features. I don't have 2003, but the differences that have
been discussed in the NGs are
- Probability functions are more accurate in 2003 (impacts NORMSDIST,
NORMDIST, LOGNORMDIST, and all ...INV functions. Also discrete
probability functions return answers over a larger range of values than
in previous versions. Text books are unlikely to have set problems
requiring the use of Excel where Excel 2002 performed poorly, but if it
should these 2003 answers will be more correct except for rare instances
where the improved inverse algorithms fail to converge.
- Statistics functions are generally more accurate in 2003 (impacts
STDEV and VAR type functions, RSQ, LINEST, TREND etc). SLOPE,
INTERCEPT, FORECAST, and STEYX will be more accurate only if there are
no missing values (see next note). Again, text books are unlikely to
have set problems where these had performed poorly, but where Excel 2003
gets a different answer, Excel 2003's answer should be better.
- Some simple linear regression functions (SLOPE, INTERCEPT, FORECAST,
and STEYX) give incorrect results when there are missing values in the
data. If the course covers statistics, it is possible that this might
be a problem. MS is working on a patch, but timing has not been disclosed.
http://groups.google.com/groups?selm=3FF963BB.2090209@no_e-mail.com
gives some suggested workarounds.
- The RAND function has a bug in it. If the course covers simulation,
then this could be a problem. MS is working on a patch, but timing has
not been disclosed.
http://groups.google.com/[email protected]
suggests an add-in with a very good random number generator.
If you would prefer 2002, you can probably get it on line from companies
that sell software remainders, or on e-bay.
Jerry