[Thuban-list] classification of variable ratios, not only single variables

[Moritz Lennert]

Daniel,

Let's keep the discussion going concerning ratios, now that the Debian
issue seems cleared up.

Daniel Calvelo Aros said:
> On Fri, 5 Mar 2004 16:42:34 +0100 (CET), Moritz Lennert wrote
>> Hello,
>>
>> I'll just keep going until someone stops me ;-)
>>
>> How difficult would it be to introduce the possibility of
>> classifying a map according to the ratio of two variables instead of
>> one variable alone. As an example: if I have a map whose data table
>> contains population and surface for each spatial entity, I have to
>> go into the data file (outside of thuban) and calculate the
>> population density and then I can generate a map of this density. It
>> would be nice to have the possibility to indicate two variables, one
>> as nominator and one as denominator and then launch the
>> classification on this ratio.
>>
>> Moritz
>
> It wouldn't be hard. The point is to do it right.
>
> I mean: the fine way would be to introduce a variable calculator into
> thuban,
> and the possibility to add new (transient) variables into a table. Since
> we
> are using SQLite, the operators are there; there is no need for a full
> expression parser.
>

Could you elaborate on this ? I am not sure I understand correctly. You
mean that instead of having to include all variables into the table, we
should be able to import (transient) variables from other sources ?


> The odd way (to me) would be to use, like ArcView does, a "Normalize by"
> extra
> variable in the classification dialog. I'd rather have the first, although
> the
> second seems more straightforward for interactive exploration.

Yes, it is very easy to use. I guess we also have to watch out concerning
the direction of Thuban: it is a viewer not a GIS, or ?

>
> Just thinking out loud, but what do you all think of having a *postgis*
> calculator, including the GEOS-based extensions, so that we could within
> thuban have choropleth maps of, say, the length of rivers that are
> contained
> within state boundaries? That's a lot of extra packages needed, and
> installing
> postgis isn't very straightforward right now...
>

And it would oblige everyone to use postgis to be able to calculate simple
ratios ? For many usages this seems overkill.

Moritz