Python for Science

Python is free, it is open source, and it has a huge community.

Python is one of the most popular and loved programming languages in the world!

The CodeEval blog published its data on the "Most Popular Coding Languages" on February 2016. It shows Python in the first place in popularity, based on usage in the CodeEval community. Meanwhile, the StackOverflow trends graph shows increasing interest in Python during the last 5 years.

Python can be used for many things: managing data bases, creating graphical user interfaces, making websites, and much more… including science. Because of the many uses, the world of Python includes many, many Libraries (you load the parts that you need).

In science, the two libraries that are king and queen of the world are: NumPy, and Matplotlib.

Numpy

NumPy is for working with data in the form of arrays (vectors, matrices). It has a myriad built-in functions or methods that work on arrays directly. To load the library into your current session of interactive Python, into a saved Python script, or into a Jupyter notebook, you use:

import numpy

Tips:

  • a one-dimensional array (vector) has the form: [1.0, 0.5, 2.5]
  • a two-dimensional array (matrix) has the form: [[ 1.0, 0.5, 2.5], [ 0.5, 1.1, 2.0]]
  • the elements in an array are numbered with an index that starts at 0
  • the colon notation: in any index position, a : means "all elements in this dimension"
  • once numpy is loaded, its built-in functions are called like this: numpy.function(arg) (where arg is the function argument: arrays to operats on, and parameters)

Try it!

In [1]:
import numpy
In [2]:
# By the way: comments in code cells start with a hash.
# here are two arrays, saved as variables x and y:
x = numpy.array([1.0, 0.5, 2.5])
y = numpy.array([[ 1.0, 0.5, 2.5], [ 0.5, 1.1, 2.0]])
In [3]:
# The print function works on arrays:
print(x)
[ 1.   0.5  2.5]
In [4]:
print(y)
[[ 1.   0.5  2.5]
 [ 0.5  1.1  2. ]]
In [5]:
numpy.shape(y)
Out[5]:
(2, 3)
In [6]:
numpy.shape(x)
Out[6]:
(3,)

Let's review what happened there. We first loaded numpy, giving us the full power to use arrays. We created two arrays: x and y… then we print x and we print y. They look nice.

Numpy has a built-in function to find out the "shape" of an array, which means: how many elements does this array have in each dimension? We find that y is a two-by-three array (it has two dimensions).

What is the first element of x? We can use square brackets and the zero-index to find out:

In [7]:
x[0]
Out[7]:
1.0

Exercise: Now, try it yourself. What is the first element of y?

Right. The first element of y is a 3-wide array of numbers. If we want to access the first element of this now, we use:

In [8]:
y[0][0]
Out[8]:
1.0

Exercise: Try picking out different elements of the array y

We learned that:

  • The square brackets allow us to pick out the elements of an array using an index: x[i]
  • For a two-dimensional array, we can use two indices: y[i][j]
  • All indices start at zero.

This is super powerful!

Matplotlib

Matplotlib is for making all kinds of plots. To get an idea of the great variety of plots possibe, have a look at the online Gallery. You can see that Matplotlib itself is a pretty big library. We can load a portion of the library (called a module) that has the basic plotting funtions with:

from matplotlib import pyplot

Once the pyplot module is loaded, its built-in functions are called like this: pyplot.function(arg) (where arg is the function argument).

An example: size of households in the US

Did you know that the size of households—that is, the number of people living in each household—has been steadily decreasing in the US and many other countries? This has perhaps surprising consequences. Even if population growth slows down, or stops altogether, the number of households keeps increasing at a fast rate.

More households means more $CO_2$ emissions! This is bad for the planet.

Get the data

Go to the folder data on the GitHub repository: Caminos, then select the file statistic_id183648.csv, click where it says "raw" and finally right-click and select Save Page As (choosing a location on your local drive).

In [9]:
#Load the data from local disk 
year, av_size = numpy.loadtxt(fname='data/statistic_id183648.csv', delimiter=',', 
                              skiprows=1, unpack=True)
In [10]:
print(year)
[ 2016.  2015.  2014.  2013.  2012.  2011.  2010.  2009.  2008.  2007.
  2006.  2005.  2004.  2003.  2002.  2001.  2000.  1999.  1998.  1997.
  1996.  1995.  1994.  1993.  1992.  1991.  1990.  1989.  1988.  1987.
  1986.  1985.  1984.  1983.  1982.  1981.  1980.  1979.  1978.  1977.
  1976.  1975.  1974.  1973.  1972.  1971.  1970.  1960.]

Exercise: Now print the variable av_size, correspondig to the average size of households (in numbers of people) for each year:

Great! The next thing we want to do is make a plot of the changing size of households over the years. To do that, we need to load the Matplotlib module called pyplot:

In [11]:
from matplotlib import pyplot
%matplotlib inline

What's this from business about? matplotlib is a pretty big (and awesome!) library. All that we need is a subset of the library for creating 2D plots, so we ask for the pyplot module of the matplotlib library.

Plotting the data is as easy as calling the function plot() from the module pyplot.

In [12]:
pyplot.plot(year, av_size)
Out[12]:
[<matplotlib.lines.Line2D at 0x10ff57978>]

But what if we'd like to get a title on this plot, or add labels to the axes? (We should always have labelled axes!). Also, we notice a long jump from the year 1960 to 1970: let's add markers to the plot and change the line style to a dotted line.

In [13]:
pyplot.plot(year, av_size, linestyle=':', marker='o')
pyplot.title("Household size in the US, 1960–2016 \n", fontsize=16)
pyplot.ylabel("Average number of people per household")
Out[13]:
<matplotlib.text.Text at 0x110238fd0>

Exercise: In the same cell above, now add a label on the x-axis, using the pyplot.xlabel() function, and re-execute it.

Python for science, so far

You learned about:

  • loading the Python libraries for science
  • using data in the form of arrays with the numpy library
  • accessing elements of an array
  • loading data from a file
  • plotting data with the Matplotlib library
  • adding title, labels to a plot, and changing the style

Data source


(c) 2017 Lorena A. Barba, Natalia Clementi. Free to use under the Creative Commons Attribution CC-BY 4.0 License. Written for the tutorial "Data Science for a Better World", at the GW _Caminos al Futuro_ Summer program.