secondspass / jupyter_bootcampproject_examples

This repo contains Jupyter and excel examples that will be useful when putting together the bootcamp projects. We will make our actual projects in different locations. This will just hold useful examples that we can all consult when we need something.

This includes examples of how to do things with Geopandas, widgets, making plots and maps.

In order to set up the environment, run conda env create -f environment.yml

Or if that is not working, pip installing the following should be sufficient pip install pandas geopandas geodatasets scipy matplotlib openpyxl

Add any data that needs to read for the examples in data. (And if github won't allow you to upload that data, then just leave some instructions in this README for your particular example

Info about each example

us_choropleth_map.ipynb

This gives a nice interactive map of the US where you can hover over counties to see the Jun average temperature and the number of medicare benificiaries in that county.

No special instructions. Just make sure that the CtyAvTemp62016.csv and 2016_HHSemPOWERMapHistoricalDataset.xlsx files are in the data directory.

aggregate_poweroutage_data.ipynb

This shows how to use Pandas groupby and aggregate functions to get mean and median (and other stuff too) of a time series data. In this example I use the eaglei_outages_2016.csv which is a time series data. I show how to aggregate the data for the date 2016-06-20 to get mean and median of the number of people who are out of power that day.

Make sure the eaglei_outages_2016.csv file is in the data directory. Github will not allow it to be pushed so you'll have to copy it in to the data directory.

Geo_project_widgets.ipynb

This shows how to apply widgets to a given Pandas/GeoPandas plot. Each Notebook cell can be run individually without depending on the others. The first cell just shows a normal GeoPandas plot. The second cell introduces a lot of widgets that update the plot (e.g., variable changing and colorbar limit changing) The third cell is a simplified version of the second cell that only allows the changing of variables

Switching variables via the dropdown box will switch the variable that is plotted (and reset the colorbar limits to default min/max values). This allows viz-ing plots for different variables very simple without having to swap code around.

Note: Sometimes the notebook widgets bug out and don't produce a plot, or it ends up producing multiple plots on accident This is usually because the kernel hiccuped when executing the notebook and usually restarting the kernel and clearing all output will help this.

Note 2: Although switching variables resets the colorbar limits to the default min/max values for that variable, currently the text in the widget itself for c_min and c_max don't update. Just a quality of life bug. A workaround can be done that fixes this by uncommenting the c_min.value and c_max.value lines in the update function, but then this generates two additional plots as updating c_min and c_max "regenerates" the plot.

eaglei_mpi_scripts

To not mess up the existing environment, these scripts only depend on installing mpi4py and pandas. Have not tested these in a notebook, and unsure how that would work but should be possible? (at the very least can have a Jupyter cell that executes a python script externally rather than the code being in the cell itself)

There are four scripts in the eaglei_mpi_scripts directory that represent two different analyses. For each analysis, there exists a non-MPI approach eaglei_no_mpi*.py and an MPI approach eaglei_mpi*.py. Each script can be run individually to see how a non-MPI approach is slow compared to when parallelizing things with MPI. Similar to other eaglei examples in this repository, you need to have access locally to eaglei_outages_*.csv for 2014-2022.

Analysis 1:

• eaglei_no_mpi.py: Sums the total blackouts across 2014-2022 without MPI. python3 eaglei_no_mpi.py
• eaglei_mpi.py: Sums the total blackouts across 2014-2022 using MPI. Each MPI rank reads in a different year. Because there are 9 files, this is run with mpirun -np 9 python3 eaglei_mpi.py

Analysis 2:

• eaglei_no_mpi_ex2.py: For a specific year (in this case 2014), sums the total blackouts for unique counties. python3 eaglei_no_mpi_ex2.py
• eaglei_mpi_ex2.py: For a specific year (in this case 2014), sums the total blackouts for unique counties using MPI. Each MPI ranks takes a subset of the counties and calculates the sum for its assigned counties. For simplicity, things are coded assuming counties divide evenly among MPI ranks, so this was tested for 1,2,4,and 8 MPI ranks (acts as a nice example of how scaling works). This is run like so (e.g., for 2 MPI ranks): mpirun -np 2 python3 eaglei_mpi_ex2.py