Gabmap has migrated to a new server! Performance should be better and there should not be any SSL errors anymore. Unfortunately, if you have old projects it may be the case that not all options work anymore. In that case please create a new project with the same files. If there are any remaining issues, please contact Martijn Wieling.
A new version of Gabmap will be installed next Wednesday (20 February 2012) between 10:00 and 12:00 (CET). The new version includes major improvements on ‘Cluster Determinants’, and some minor bug fixes. Please do not hesitate to contact if you experience any problems. If you’d like to try the new version, you can already try it at http://gabmap.nl/testing (the data is not shared with the main site, you will need to create a new account).
Because of the large amount of spam, comments for all posts and pages are now disabled for non-subscribers. See About for contacting by e-mail.
The scoring method in the Gabmap part cluster determinants has changed. Please report any adverse effects on performance caused by this change.
It is now possible to use kml files that contain MultiGeometry. If you still notice problems with these type of kml files, please let me know.
Now available, the final version of:
John Nerbonne, Rinke Colen, Charlotte Gooskens, Peter Kleiweg, and Therese Leinonen (2011). Gabmap — A Web Application for Dialectology
The processing of numeric data has changed. This change does not effect existing projects.
There were two methods to choose from to join the results of individual columns, Euclidean Distance and Manhattan Distance. There is now only one method, which is average distance.
Euclidean Distance emphasises larger values over smaller ones. That doesn’t make much sense in the current set-up.
Manhattan Distance is just the unscaled version of average distance.
The use of Cronbach’s Alpha has also changed. It used to be applied to the numeric data itself. Now it is applied to the measurements of the data, just like it is with the processing of string data and categorical data.