Python instructions for linux (mint/ubuntu) about geospatialstroke HOT 8 CLOSED

I'm facing related issues; but a few steps before trying to get jupyter notebooks to run

When you come into the office early to get some non-work work done, you spend 50 minutes getting python and jupyter to (re)install, you forget what task you wanted to do in the first place... #python

— Symbolix (@SymbolixAU) February 11, 2019

from geospatialstroke.

@richardbeare your pyproj installation is misconfigured. I haven't used pip to install the stack in a few years so it's a bit hard to guess what's happening with that install. In general I'd strongly suggest conda + conda-forge for easy management of the python spatial data science stack. On linux, something like:

wget http://bit.ly/miniconda -O miniconda.sh
bash miniconda.sh -b -p $HOME/miniconda
export PATH="$HOME/miniconda/bin:$PATH"
hash -r
conda update conda
conda config --add channels conda-forge --force
conda create --name my_env python=3 --file requirements.txt
source activate my_env
jupyter lab

from geospatialstroke.

OK,
I haven't played with conda - will look into it. We'll need some introduction about the appropriate environments for both R and python in our supplementary material.

from geospatialstroke.

@gboeing Is it possible to render the ipynb to .html (or .md) ?

from geospatialstroke.

Yep! Gonna put my son down for bed then return to this in ~30 minutes.

from geospatialstroke.

The conda instructions worked perfectly.

Is it possible to display leaflet map in the notebook?

In the last map, colour should be by stroke count:

folium.Choropleth(gdf_nearby, data=gdf_nearby, columns=['POA_CODE', 'strokes'],
                  key_on='feature.properties.POA_CODE',
                  legend_name='Number of strokes', fill_color='YlOrRd').add_to(m)

from geospatialstroke.

I'm working through example 2 - it runs perfectly.

Query on the voronoi tesselation - was the basin expansion step required due to keeping the largest polygon in the previous cell?

Also, the final stroke calculation seems to be different to what we're doing on the R side, if I'm reading it correctly. My thought was to use the proportions of simulated cases from a postcode to drive the calculation. It seems to me that any postcode intersecting a catchment is included, rather than a fraction of a postcode? i.e. we should assign the 1000 sample locations to the containing postcodes, then compute a proportion from each postcode attending each destination, then multiply that by the predicted cases, then add those up.

There are certainly some large areas of nothing in that part of Melbourne (airports and former rural areas that are currently being developed), which could bias calculations in strange ways.

from geospatialstroke.

@richardbeare @SymbolixAU

We'll need some introduction about the appropriate environments for both R and python in our supplementary material.

I'll add this to the subdirectory's readme.

Is it possible to render the ipynb to .html (or .md) ?

I'll include .html versions in the notebooks folder.

Is it possible to display leaflet map in the notebook?

Yes, and they were in an earlier commit, but I removed them because it adds a lot of size to the commits and in turn the repo. Your call on if it's in the notebook directly or not.

In the last map, colour should be by stroke count

I'll adjust accordingly.

was the basin expansion step required due to keeping the largest polygon in the previous cell?

It was because the tessellation left parts of the periphery patchy. As the sample size -> infinity, the patches will disappear, but with a moderate n I wanted to make sure to fill in all the gaps for nice complete basins.

It seems to me that any postcode intersecting a catchment is included, rather than a fraction of a postcode?

I'll adjust accordingly.

from geospatialstroke.