These weeks several problems and clarifications have been made on the
"road" to release finally EpiData Analysis as v2.0
Some aspects are still insufficiently final, but the level of changes
are going into "corner" details from my point of view.
As of build 122 some aspects have been fixed:
e.g.
- EpiDemiological tables (odds ratios, rr and attributable risk) will
use highest value of case and exposure variables as default.
- A new table dialog has been a success in my personal teaching. Quite
clearly beginners adapt much faster when having proper (intuitive) dialogs.
- Sorting of frequency tables by numbers
- More graph types can now be used with a "group variable", e.g.: box
age /by=sex
- Many more subtle details have been fixed. Many of these are invisible
to the end-user, but clarification makes the whole programme more
stable. This will also allow shorter development time for future new
commands and additions.
Some issues are still creating problems, such as:
- internal errors after kruskal-wallis tests.
- parts of relate system not working for some files
- creating more intuitive dialogs for remaining parts of the system
- removing reported bugs (see mantis system).
Some users might find it odd that new problems arise, but this is due to
the rewriting of core parts. And consequential changes in other parts of
the software take some time to find, if they are not pinpointed by the
error tracking system. This system is a collection of pgm files doing
all verifying results and function in all parts of the programme.
Usually when a new error is reported it is due to not being tested for
in the error tracking system. The system is now based on 41 pgm files
with around 1150 different conditions being tested and verified.
Version 2 will not be released before all of the consequences of the
rewriting of core parts has taken place. But as could be stated on the
www.epidata.dk/testing.php page : "Many of the functions are running and
functioning very well, so after each user has verified the function of
what he/she uses there is no problem in using the test versions." E.g.
the table module has no known errors just now.
Kind regards
Jens Lauritsen
EpiData Association