I am currently trying to reconstruct more or less convincing networks for the north-eastern part of Germany based on the EU-DEM using the tool r.walk in GRASS GIS. After showing talking about that during the last meeting, I tried out a data set of a more hilly region (around the nice mountain Brocken in central Germany).
I ran a few tests with several different start and stop points, trying to reconstruct a network between some of the larger settlements surrounding the mountain. The workflow was basically 1) r.walk with 1 of the points as the start and all the others as the targets iterated for all points and 2) r.drain on the resulting cost surfaces inverted, meaning the start point of r.walk is now the target point and vice versa. What I got is something like this.
My questions would be:
1) Is there a way to script this? I did not find a way to convince QGIS or GRASS GIS to select every point in a layer in an iterative process for the start points for r.walk and to invert this selection for the corresponding target points.
2) Since r.walk is nonisotropic, sometimes there are two paths connecting two points, according to which direction one walks. Is there a way to normalize this?
This is a brief explanation of the problem I presented in the 28. January 2020 GIS stammtisch.
I have a database in Excel. the larfer part of the data is nominal. The solumns represent categorical variables and the rows represent the sites.
I am looking for different methods that I can analyze the data. group the sites, or find correlations between variables.
I would like to find groups of sites that are more similar to eachother according to this data.
In total I have around 60 sites and 30 variables.
I created a contingency table and tried the Chi square test.
I also converterd the data into binary data and made the fisher exact test on it. but it is difficult to interpret the data.
I was thinking about doing other tests like uncertainty coefficient, etc.
I am looking for other suggestions. 
