Python package exercise

Python package exercise#

In this exercises you will create your own ts_emergency package. You will first recreate the package covered in the packaging chapter and then modify the packaging module to be a namespace that contains two submodules. solutions

Exercise 1#

The first job is to create the skeleton of the ts_emergency package. A link to the example data ed_ts_mth.csv is provided below.

Task

  • Create the directory, data and python module structure below. No code need be included at this stage.

ts_emergency
├── __init__.py
├── plotting.py
├── datasets.py
├── data
│   ├── syn_ts_ed_long.csv
│   ├── syn_ts_ed_wide.csv

Data files:

The dataset syn_ts_ed_long.csv contains data from 4 emergency departments in 2014. The data are stored in long (sometimes called tidy) format. You are provided with three columns: date (non unique date time formatted), hosp (int 1-4) and attends (int, daily number of attends at hosp \(i\))

The dataset syn_ts_wide.csv contains the same data in wide format. Each row now represents a unique date and each hospital ED has its own column.

  • The ed data is held in long format here

  • The ed data is held in wide format here

Hints:

  • Remember to think about where the local package needs to be stored relative to the code that is going to use it.

  • You can choose to use either the long format or short format data for this exercise. For basic plotting is is often easier to use a wide format.

Exercise 2:#

Task:

  • Add appropraite __version__ and __author__ attributes to __init__.py

  • Check these work by importing your package and printing the relevant attributes.

Hints:

  • These should be of type str

# your code testing your package here ...

Exercise 3:#

Now that you have a structure you can add code to the modules.

Check the matplotlib exercises and solutions for help with these functions and/or the github repo for a complete solution

Task:

  • Create the following skeleton functions in the modules listed. Feel free to add your own parameters.

    • ts_emergency.datasets:

      • load_ed_ts(): returns a pandas.Dataframe or numpy.ndarray (or both via a parameter)

    • ts_emergency.plotting:

      • plot_single_ed(pandas.Dataframe, str). Simple plot of a selected time series over time.

      • Returns a matplotlib figure and axis objects

      • plot_eds(pandas.Dataframe): grid plot of all ED time series

  • test importing the functions to your code (e.g. Jupyter notebook or script).

Hint:

  • A skeleton function might look like the following:

def skeleton_example():
    pass


def skeleton_example():
    print('you called skeleton_example()')
    return None

  • importing should look like:

from ts_emergency.plotting import plot_single_ed, plot_eds
from ts_emergency.datasets import load_ed_ts

# your code testing your package here ...

Exercise 4:#

Task:

  • Complete the code for the plotting and dataset skeleton functions you have created.

  • Test your package. For example

    • Load the example ED dataset

    • Create plots of all ED time series and individual time series.

Hints

Exercise 5#

Let’s create a new major version of the package that extends the basic ts_emergency package so that it also has some simple time series analysis functionality. We class this as a major change as we will won’t be keeping backwards compatability with the current version of ts_emergency.

You will now create a plotting namespace that contains two submodules: view and tsa. The module ts_emergency.plotting.view will contain the code currently held in ts_emergency.plotting while ts_emergency.plotting.tsa will contain new functions related to plotting the results of three simple time series analysis operations.

Task

  • Create a new major version of the ts_emergency package. Update the version number of the package (e.g. to 1.0.0 or 2.0.0 depending on your initial version choice).

  • The new package should have the structure below.

    • A key change is that plotting is now a directory.

    • The view module is the old plotting module. Just rename it.

    • tsa is a new module

    • It is important to include ts_emergency/plotting/__init__.py. This allows us to treat ts_emergency/plotting as a namespace (that contains submodules).

ts_emergency
├── __init__.py
├── plotting
│   ├── __init__.py
│   ├── view.py
│   ├── tsa.py
├── datasets.py
├── data
│   ├── syn_ts_ed_long.csv
│   ├── syn_ts_ed_wide.csv

  • Test the view module by importing plot_single_ed()

Hints:

  • There is no need to include a __version__ in ts_emergency/plotting/__init__.py.

# your code testing your package here ...

Exercise 6#

You will now create two example functions for tsa. This exercise also provides some matplotlib practice and a brief introduction to statsmodels time series analysis functionality.

plot_detrended

For a given ED time series, this function generates and plots a differenced or detrended ED time series. The 1st difference is the difference between \(y_{t+1}\) and \(y_t\).

The output of the function should be a plot similar to the below. The function could return the fig and ax objects for a user.

detrended

diagnostics_plot

For a given ED time series, the function will generate a plot similar to the below:

diag

The figure consists of three axis objects. The first plots the detrended series. The second plot is the autocorrelation function (ACF): a measure of correlation of a variable with previous observations of itself. The third is the partial autocorrelation function (PACF): a measure of correlation of a variable with early observations of itself while controlling (regressing) for shorter lags. The good news is you can create ACF and PACF using two functions from statsmodels

#  import the functions
from statsmodels.graphics.tsaplots import plot_acf, plot_pacf

Task:

  • Code the plot_detrended and diagnostics_plot functions and add them to the ts_emergency/plotting/tsa module.

Hints:

  • diagnostics_plot is a good test of your matplotlib skills!

  • Try creating each plot indipendently first.

  • Note that the plot_acf and plot_pacf accepts a ax parameter. Can you use this parameter? to add the plot to the correct place?

  • There are various ways to answer this question. Consider using a gridspec.

  • Check out documentation for plot_acf and plot_pacf on the statsmodels docs. For example

# your package testing code here ...

Exercise 7:#

Task

  • Think about python programmes you have coded in the past. Can you think of how you would organise them as packages i.e. package name, submodules and example data? Choose a suitable example and draft an outline the structure of the package.

Contents