Hi again, The scipy version in osgeo4w is 0.11.0 [1], so I guess we can simply assume that gaussian_kde.set_bandwidth exists, right?
Regarding the LSCV implementation, I'll ask in the scipy mailing list. Finally, I asked Víctor Olaya about the algorithm definition and he says that it's possible to create the different raster outputs, but it's not possible to define them beforehand. This means that the outputs can be created and added as layers but since they are not defined, it may cause some issues, for example, in the graphic modeler. Is that a problem or should we simply create the outputs without previous definition? Cheers, Víctor. [1] http://download.osgeo.org/osgeo4w/versions.html 2013/5/2 Paolo Cavallini <[email protected]> > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Il 02/05/2013 11:28, Victor Gonzalez ha scritto: > > > We already started analyzing the task and we have some questions. First, > we are going > > to made a fork of the current animove plugin repository [1] and when we > finish we'll > > do a pull request. Is that ok with you? Anyone has a better idea? > > I think you could take the lead, no need for the pull. Francesco? > > > Regarding the kernel analysis algorithm, we found that the gaussian_kde > class in > > scipy [2] has a set_bandwidth method in order to set the bandwidth of > the method as a > > simple scalar. But this is only available from 0.11.0 version. Which > version of scipy > > we assume will be installed in the distribution? > > Please check the one available in osgeo4w. In Debian we have > 0.10.1+dfsg2-1. > Maybe you can implement it conditionally, so that it will be available > when scipy > will be upgraded. > > > Also, that gaussian_kde class allows 'scotts' and 'silverman' method to > calculate the > > bandwidth, but no LSCV or is available. The only way to have it working > is to set the > > bw_method as a callable and implement the LSCV bandwidth ourselves. The > problem is > > that we are not experts in statistics and we don't know how to implement > it. We > > searched for the formula but it's quite hard to understand and implement > at first. Do > > you know of any python implementation? > > I would avoid implementing stuff locally, unless it's a relatively simple > extension > to numpy/scipy. Always better to try to find another py library, or > cooperate > upstream with scipy devs. > > > Finally, the kernel analysis is supposed to create several probability > rasters (one > > for each animal), right? The problem is that we don't know the number of > outputs > > beforehand, so we cannot specify them in the SEXTANTE algorithm. A > solution is to > > create the rasters in a folder (specified by the user) and then add it > to QGIS by > > hand. Is that ok? > > Partly: I think there should be a way to add them all. Please check with > Victor > Olaya, it would be better to add the function to Sextante core if > necessary. > > All the best. > - -- > Paolo Cavallini - Faunalia > www.faunalia.eu > Full contact details at www.faunalia.eu/pc > Nuovi corsi QGIS e PostGIS: http://www.faunalia.it/calendario > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.4.12 (GNU/Linux) > Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ > > iEYEARECAAYFAlGCN3gACgkQ/NedwLUzIr4ZKwCfRKtqQICDLbPbPKqLhMzCwQ3X > 5sEAnjZLTR7Tkk4QP9UtyB6Z2naM5PiP > =tmjK > -----END PGP SIGNATURE----- > _______________________________________________ > AniMov mailing list > [email protected] > http://lists.faunalia.it/cgi-bin/mailman/listinfo/animov >
_______________________________________________ AniMov mailing list [email protected] http://lists.faunalia.it/cgi-bin/mailman/listinfo/animov
