Commit 24882ffa authored by Michiel Cottaar's avatar Michiel Cottaar
Browse files

Merge branch 'master' into 'master'

Several minor additions to the talk (including pandas-profiling)

See merge request fsl/pytreat-practicals-2020!28
parents 04f898f4 b802870a
%% Cell type:markdown id: tags:
# Pandas
Follow along online at: https://git.fmrib.ox.ac.uk/fsl/pytreat-practicals-2020/-/blob/master/talks/pandas/pandas.ipynb
Pandas is a data analysis library focused on the cleaning and exploration of
tabular data.
Some useful links are:
- [main website](https://pandas.pydata.org)
- [documentation](http://pandas.pydata.org/pandas-docs/stable/)<sup>1</sup>
- [Python Data Science Handbook](https://jakevdp.github.io/PythonDataScienceHandbook/)<sup>1</sup> by
Jake van der Plas
- [List of Pandas tutorials](https://pandas.pydata.org/pandas-docs/stable/getting_started/tutorials.html)
<sup>1</sup> This tutorial borrows heavily from the pandas documentation and
the Python Data Science Handbook
%% Cell type:code id: tags:
```
%pylab inline
import pandas as pd # pd is the usual abbreviation for pandas
import matplotlib.pyplot as plt # matplotlib for plotting
import seaborn as sns # seaborn is the main plotting library for Pandas
import statsmodels.api as sm # statsmodels fits linear models to pandas data
import statsmodels.formula.api as smf
from IPython.display import Image
sns.set() # use the prettier seaborn plotting settings rather than the default matplotlib one
```
%% Cell type:markdown id: tags:
> We will mostly be using `seaborn` instead of `matplotlib` for
> visualisation. But `seaborn` is actually an extension to `matplotlib`, so we
> are still using the latter under the hood.
## Loading in data
Pandas supports a wide range of I/O tools to load from text files, binary files,
and SQL databases. You can find a table with all formats
[here](http://pandas.pydata.org/pandas-docs/stable/io.html).
%% Cell type:code id: tags:
```
titanic = pd.read_csv('https://raw.githubusercontent.com/mwaskom/seaborn-data/master/titanic.csv')
titanic
```
%% Cell type:markdown id: tags:
This loads the data into a
[`DataFrame`](https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.html)
object, which is the main object we will be interacting with in pandas. It
represents a table of data. The other file formats all start with
`pd.read_{format}`. Note that we can provide the URL to the dataset, rather
than download it beforehand.
We can write out the dataset using `dataframe.to_{format}(<filename)`:
We can write out the dataset using `dataframe.to_{format}(<filename>)`:
%% Cell type:code id: tags:
```
titanic.to_csv('titanic_copy.csv', index=False) # we set index to False to prevent pandas from storing the row names
```
%% Cell type:markdown id: tags:
If you can not connect to the internet, you can run the command below to load
this locally stored titanic dataset
%% Cell type:code id: tags:
```
titanic = pd.read_csv('titanic.csv')
titanic
```
%% Cell type:markdown id: tags:
Note that the titanic dataset was also available to us as one of the standard
datasets included with seaborn. We could load it from there using
%% Cell type:code id: tags:
```
sns.load_dataset('titanic')
```
%% Cell type:markdown id: tags:
`Dataframes` can also be created from other python objects, using
`pd.DataFrame.from_{other type}`. The most useful of these is `from_dict`,
which converts a mapping of the columns to a pandas `DataFrame` (i.e., table).
%% Cell type:code id: tags:
```
pd.DataFrame.from_dict({
'random numbers': np.random.rand(5),
'sequence (int)': np.arange(5),
'sequence (float)': np.linspace(0, 5, 5),
'letters': list('abcde'),
'constant_value': 'same_value'
})
```
%% Cell type:markdown id: tags:
For many applications (e.g., ICA, machine learning input) you might want to
extract your data as a numpy array. The underlying numpy array can be accessed
using the `values` attribute
using the `to_numpy` method
%% Cell type:code id: tags:
```
titanic.values
titanic.to_numpy()
```
%% Cell type:markdown id: tags:
Note that the type of the returned array is the most common type (in this case
object). If you just want the numeric parts of the table you can use
`select_dtypes`, which selects specific columns based on their dtype:
%% Cell type:code id: tags:
```
titanic.select_dtypes(include=np.number).values
titanic.select_dtypes(include=np.number).to_numpy()
```
%% Cell type:markdown id: tags:
Note that the numpy array has no information on the column names or row indices.
Alternatively, when you want to include the categorical variables in your later
analysis (e.g., for machine learning), you can extract dummy variables using:
%% Cell type:code id: tags:
```
pd.get_dummies(titanic)
```
%% Cell type:markdown id: tags:
## Accessing parts of the data
[Documentation on indexing](http://pandas.pydata.org/pandas-docs/stable/indexing.html)
### Selecting columns by name
Single columns can be selected using the normal python indexing:
%% Cell type:code id: tags:
```
titanic['embark_town']
```
%% Cell type:markdown id: tags:
If the column names are simple strings (not required) we can also access it
directly as an attribute
%% Cell type:code id: tags:
```
titanic.embark_town
```
%% Cell type:markdown id: tags:
Note that this returns a pandas
[`Series`](https://pandas.pydata.org/pandas-docs/stable/generated/pandas.Series.html)
rather than a `DataFrame` object. A `Series` is simply a 1-dimensional array
representing a single column. Multiple columns can be returned by providing a
list of columns names. This will return a `DataFrame`:
%% Cell type:code id: tags:
```
titanic[['class', 'alive']]
```
%% Cell type:markdown id: tags:
Note that you have to provide a list here (square brackets). If you provide a
tuple (round brackets) pandas will think you are trying to access a single
column that has that tuple as a name:
%% Cell type:code id: tags:
```
titanic[('class', 'alive')]
```
%% Cell type:markdown id: tags:
In this case there is no column called `('class', 'alive')` leading to an
error. Later on we will see some uses to having columns named like this.
### Indexing rows by name or integer
Individual rows can be accessed based on their name (i.e., the index) or integer
(i.e., which row it is in). In our current table this will give the same
results. To ensure that these are different, let's sort our titanic dataset
based on the passenger fare:
%% Cell type:code id: tags:
```
titanic_sorted = titanic.sort_values('fare')
titanic_sorted
```
%% Cell type:markdown id: tags:
Note that the re-sorting did not change the values in the index (i.e., left-most
column).
We can select the first row of this newly sorted table using `iloc`
%% Cell type:code id: tags:
```
titanic_sorted.iloc[0]
```
%% Cell type:markdown id: tags:
We can select the row with the index 0 using
%% Cell type:code id: tags:
```
titanic_sorted.loc[0]
```
%% Cell type:markdown id: tags:
Note that this gives the same passenger as the first row of the initial table
before sorting
%% Cell type:code id: tags:
```
titanic.iloc[0]
```
%% Cell type:markdown id: tags:
Another common way to access the first or last N rows of a table is using the
head/tail methods
%% Cell type:code id: tags:
```
titanic_sorted.head(3)
```
%% Cell type:code id: tags:
```
titanic_sorted.tail(3)
```
%% Cell type:markdown id: tags:
Note that nearly all methods in pandas return a new `Dataframe`, which means
that we can easily call another method on them
%% Cell type:code id: tags:
```
titanic_sorted.tail(10).head(5) # select the first 5 of the last 10 passengers in the database
```
%% Cell type:code id: tags:
```
titanic_sorted.iloc[-10:-5] # alternative way to get the same passengers
```
%% Cell type:markdown id: tags:
**Exercise**: use sorting and tail/head or indexing to find the 10 youngest
passengers on the titanic. Try to do this on a single line by chaining calls
to the titanic `DataFrame` object
%% Cell type:code id: tags:
```
titanic.sort_values...
```
%% Cell type:markdown id: tags:
### Indexing rows by value
One final way to select specific columns is by their value
%% Cell type:code id: tags:
```
titanic[titanic.sex == 'female'] # selects all females
```
%% Cell type:code id: tags:
```
# select all passengers older than 60 who departed from Southampton
titanic[(titanic.age > 60) & (titanic['embark_town'] == 'Southampton')]
```
%% Cell type:markdown id: tags:
Note that this required typing `titanic` quite often. A quicker way to get the
same result is using the `query` method, which is described in detail
[here](http://pandas.pydata.org/pandas-docs/stable/indexing.html#the-query-method)
(note that using the `query` method is also faster and uses a lot less
memory).
> You may have trouble using the `query` method with columns which have
a name that cannot be used as a Python identifier.
%% Cell type:code id: tags:
```
titanic.query('(age > 60) & (embark_town == "Southampton")')
```
%% Cell type:markdown id: tags:
When selecting a categorical multiple options from a categorical values you
might want to use `isin`:
%% Cell type:code id: tags:
```
titanic[titanic['class'].isin(['First','Second'])]
```
%% Cell type:markdown id: tags:
Particularly useful when selecting data like this is the `isna` method which
finds all missing data
%% Cell type:code id: tags:
```
titanic[~titanic.age.isna()] # select first few passengers whose age is not N/A
```
%% Cell type:markdown id: tags:
This removing of missing numbers is so common that it has is own method
%% Cell type:code id: tags:
```
titanic.dropna() # drops all passengers that have some datapoint missing
```
%% Cell type:code id: tags:
```
titanic.dropna(subset=['age', 'fare']) # Only drop passengers with missing ages or fares
```
%% Cell type:markdown id: tags:
**Exercise**: use sorting, indexing by value, `dropna` and `tail`/`head` or
indexing to find the 10 oldest female passengers on the titanic. Try to do
this on a single line by chaining calls to the titanic `DataFrame` object
%% Cell type:code id: tags:
```
titanic...
```
%% Cell type:markdown id: tags:
## Plotting the data
Before we start analyzing the data, let's play around with visualizing it.
Pandas does have some basic built-in plotting options:
%% Cell type:code id: tags:
```
titanic.fare.hist(bins=20, log=True)
```
%% Cell type:code id: tags:
```
titanic.age.plot()
```
%% Cell type:markdown id: tags:
Individual columns are essentially 1D arrays, so we can use them as such in
To plot all variables simply call `plot` or `hist` on the full dataframe
rather than a single Series (i.e., column). You might want to set `subplots=True`
to plot each variable in a different subplot.
Individual Series are essentially 1D arrays, so we can use them as such in
`matplotlib`
%% Cell type:code id: tags:
```
plt.scatter(titanic.age, titanic.fare)
```
%% Cell type:markdown id: tags:
However, for most purposes much nicer plots can be obtained using
[Seaborn](https://seaborn.pydata.org). Seaborn has support to produce plots
showing the
[univariate](https://seaborn.pydata.org/tutorial/distributions.html#plotting-univariate-distributions)
or
[bivariate](https://seaborn.pydata.org/tutorial/distributions.html#plotting-bivariate-distributions)
distribution of data in a single or a grid of plots. Most of the seaborn
plotting functions expect to get a pandas `DataFrame` (although they will work
with Numpy arrays as well). So we can plot age vs. fare like:
%% Cell type:code id: tags:
```
sns.jointplot('age', 'fare', data=titanic)
```
%% Cell type:markdown id: tags:
**Exercise**: check the documentation from `sns.jointplot` (hover the mouse
over the text `jointplot` and press shift-tab) to find out how to turn the
scatter plot into a density (kde) map
%% Cell type:code id: tags:
```
sns.jointplot('age', 'fare', data=titanic, ...)
```
%% Cell type:markdown id: tags:
Here is just a brief example of how we can use multiple columns to illustrate
the data in more detail
%% Cell type:code id: tags:
```
sns.relplot(x='age', y='fare', col='class', hue='sex', data=titanic,
col_order=('First', 'Second', 'Third'))
```
%% Cell type:markdown id: tags:
**Exercise**: Split the plot above into two rows with the first row including
the passengers who survived and the second row those who did not (you might
have to check the documentation again by using shift-tab while overing the
mouse over `relplot`)
%% Cell type:code id: tags:
```
sns.relplot(x='age', y='fare', col='class', hue='sex', data=titanic,
col_order=('First', 'Second', 'Third')...)
```
%% Cell type:markdown id: tags:
One of the nice thing of Seaborn is how easy it is to update how these plots
look. You can read more about that
[here](https://seaborn.pydata.org/tutorial/aesthetics.html). For example, to
increase the font size to get a plot more approriate for a talk, you can use:
%% Cell type:code id: tags:
```
sns.set_context('talk')
sns.violinplot(x='class', y='age', hue='sex', data=titanic, split=True,
order=('First', 'Second', 'Third'))
```
%% Cell type:markdown id: tags:
## Summarizing the data (mean, std, etc.)
There are a large number of built-in methods to summarize the observations in
a Pandas `DataFrame`. Most of these will return a `Series` with the columns
names as index:
%% Cell type:code id: tags:
```
titanic.mean()
```
%% Cell type:code id: tags:
```
titanic.quantile(0.75)
```
%% Cell type:markdown id: tags:
One very useful one is `describe`, which gives an overview of many common
summary measures
%% Cell type:code id: tags:
```
titanic.describe()
```
%% Cell type:markdown id: tags:
For a more detailed exploration of the data, you might want to check
[pandas_profiliing](https://pandas-profiling.github.io/pandas-profiling/docs/)
(not installed in fslpython, so the following will not run in fslpython):
%% Cell type:code id: tags:
```
from pandas_profiling import ProfileReport
profile = ProfileReport(titanic, title='Titanic Report', html={'style':{'full_width':True}})
profile.to_widgets()
```
%% Cell type:markdown id: tags:
Note that non-numeric columns are ignored when summarizing data in this way.
We can also define our own functions to apply to the columns (in this case we
have to explicitly set the data types).
%% Cell type:code id: tags:
```
def mad(series):
"""
Computes the median absolute deviatation (MAD)
This is a outlier-resistant measure of the standard deviation
"""
no_nan = series.dropna()
return np.median(abs(no_nan - np.nanmedian(no_nan)))
titanic.select_dtypes(np.number).apply(mad)
```
%% Cell type:markdown id: tags:
We can also provide multiple functions to the `apply` method (note that
functions can be provided as strings)
%% Cell type:code id: tags:
```
titanic.select_dtypes(np.number).apply(['mean', np.median, np.std, mad])
```
%% Cell type:markdown id: tags:
### Grouping by
One of the more powerful features of is `groupby`, which splits the dataset on
a categorical variable. The book contains a clear tutorial on that feature
[here](https://jakevdp.github.io/PythonDataScienceHandbook/03.08-aggregation-and-grouping.html). You
can check the pandas documentation
[here](http://pandas.pydata.org/pandas-docs/stable/groupby.html) for a more
formal introduction. One simple use is just to put it into a loop
%% Cell type:code id: tags:
```
for cls, part_table in titanic.groupby('class'):
print(f'Mean fare in {cls.lower()} class: {part_table.fare.mean()}')
```
%% Cell type:markdown id: tags:
However, it is more often combined with one of the aggregation functions
discussed above as illustrated in this figure from the [Python data science
handbook](https://jakevdp.github.io/PythonDataScienceHandbook/06.00-figure-code.html#Split-Apply-Combine)
![group by image](group_by.png)
%% Cell type:code id: tags:
```
titanic.groupby('class').mean()
```
%% Cell type:markdown id: tags:
We can also group by multiple variables at once
%% Cell type:code id: tags:
```
titanic.groupby(['class', 'survived']).mean() # as always in pandas supply multiple column names as lists, not tuples
```
%% Cell type:markdown id: tags:
When grouping it can help to use the `cut` method to split a continuous variable
into a categorical one
%% Cell type:code id: tags:
```
titanic.groupby(['class', pd.cut(titanic.age, bins=(0, 18, 50, np.inf))]).mean()
```
%% Cell type:markdown id: tags:
We can use the `aggregate` method to apply a different function to each series
%% Cell type:code id: tags:
```
titanic.groupby(['class', 'survived']).aggregate((np.median, mad))
```
%% Cell type:markdown id: tags:
Note that both the index (on the left) and the column names (on the top) now
have multiple levels. Such a multi-level index is referred to as `MultiIndex`.
This does complicate selecting specific columns/rows. You can read more of using
`MultiIndex` [here](http://pandas.pydata.org/pandas-docs/stable/advanced.html).
The short version is that columns can be selected using direct indexing (as
discussed above)
%% Cell type:code id: tags:
```
df_full = titanic.groupby(['class', 'survived']).aggregate((np.median, mad))
```
%% Cell type:code id: tags:
```
df_full[('age', 'median')] # selects median age column; note that the round brackets are optional
```
%% Cell type:code id: tags:
```
df_full['age'] # selects both age columns
```
%% Cell type:markdown id: tags:
Remember that indexing based on the index was done through `loc`. The rest is