# 2 Every picture tells a story¶

It can be difficult and confusing to look at a table of rows of numbers and make any meaningful interpretation especially if there are many rows and columns.

Handily, pandas has a method called plot() which will visualise data for us by producing a chart.

Before using the plot() method, the following line of code must be executed (once) which tells Jupyter to display all charts inside this notebook, immediately after each call to plot():

In []:

%matplotlib inline

To plot ‘Max Wind SpeedKm/h ’, it’s as simple as this code:

In []:

london['Max Wind SpeedKm/h'].plot(grid=True)

Out[]:

Figure 10

Chart of the values in the Max Wind SpeedKm/h column of the london dataframe. The grid=True argument makes the gridlines (the dotted lines in the image above) appear, which make values easier to read on the chart. The chart comes out a bit small, so you can make it bigger by giving the plot() method some extra information. The figsize units are inches.

In []:

london['Max Wind SpeedKm/h'].plot(grid=True, figsize=(10,5))

Out[]:

Figure 11

Larger version of the first chart on this page That’s better! The argument given to the plot() method, figsize=(10,5) simply tells plot() that the x-axis should be 10 units wide and the y-axis should be 5 units high. In the above graph the x-axis (the numbers at the bottom) shows the dataframe’s index, so 0 is 1 January and 50 is 18 February.

The y-axis (the numbers on the side) shows the range of wind speed in kilometres per hour. It is clear that the windiest day in 2014 was somewhere in mid-February and the wind reached about 66 kilometers per hour.

By default, the plot() method will try to generate a line, although as you’ll see in a later week, it can produce other chart types too.

## Exercise 5 Every picture tells a story¶

### Question¶

Now try Exercise 5 in the Exercise notebook 2.

If you’re using Anaconda, remember that to open the notebook you’ll need to navigate to the notebook using Jupyter.

#### 2.1 Changing a dataframe’s index¶

We have seen that by default every dataframe has an integer index for its rows which starts from 0.

The dataframe we’ve been using, london , has an index that goes from 0 to 364. The row indexed by 0 holds data for the first day of the year and the row indexed by 364 holds data for the last day of the year. However, the column 'GMT' holds datetime64 values which would make a more intuitive index.

Changing the index to datetime64 values is as easy as assigning to the dataframe’s index attribute the contents of the 'GMT' column, like this:

In []:

[ ]:


london.index = london['GMT']
#Display the first 2 rows


Out[]:

Figure 12

First 2 rows of the london dataframe showing that the index has been changed to the datetime64 values from the GMT column. Note that only the first few columns are shown due to the limitation of page width. * Note that the right of the table has been cropped to fit on the page. *

Notice that the 'GMT' column still remains and that the index has been labelled to show that it has been derived from the 'GMT' column.

You can still access a row using the iloc attribute, so to get the first line in the dataframe you can simply execute:

In []:

london.iloc[0]

Out[]:

[ ]:


GMT 2014-01-01 00:00:00
Max TemperatureC 11
Mean TemperatureC 8
Min TemperatureC 6
Dew PointC 9
MeanDew PointC 7
Min DewpointC 4
Max Humidity 94
Mean Humidity 86
Min Humidity 73
Max Sea Level PressurehPa 1002
Mean Sea Level PressurehPa 993
Min Sea Level PressurehPa 984
Max VisibilityKm 31
Mean VisibilityKm 11
Min VisibilitykM 2
Max Wind SpeedKm/h 40
Mean Wind SpeedKm/h 26
Max Gust SpeedKm/h 66
Precipitationmm 9.91
CloudCover 4
Events Rain
WindDirDegrees 186
Name: 2014-01-01 00:00:00, dtype: object


But now you can now also use the datetime64 index to get a row using the dataframe’s loc attribute, like this:

In []:

london.loc[datetime(2014, 1, 1)]

Out[]:

[ ]:


GMT 2014-01-01 00:00:00
Max TemperatureC 11
Mean TemperatureC 8
Min TemperatureC 6
Dew PointC 9
MeanDew PointC 7
Min DewpointC 4
Max Humidity 94
Mean Humidity 86
Min Humidity 73
Max Sea Level PressurehPa 1002
Mean Sea Level PressurehPa 993
Min Sea Level PressurehPa 984
Max VisibilityKm 31
Mean VisibilityKm 11
Min VisibilitykM 2
Max Wind SpeedKm/h 40
Mean Wind SpeedKm/h 26
Max Gust SpeedKm/h 66
Precipitationmm 9.91
CloudCover 4
Events Rain
WindDirDegrees 186
Name: 2014-01-01 00:00:00, dtype: object


A query such as ‘Return all the rows where the date is between 8 December and 12 December’ which you did before (and can still do) with:

In []:

[ ]:


london[(london['GMT'] >= datetime(2014, 12, 8))
& (london['GMT'] <= datetime(2014, 12, 12))]


can now be done more succinctly like this:

In []:

[ ]:



london.loc[datetime(2014,12,8) : datetime(2014,12,12)]

#The meaning of the above code is get the rows between

#and including the indices datetime(2014,12,8) and
#datetime(2014,12,12)


Out[]:

Figure 13

Rows from the london dataframe where the index is between 2014-12-08 and 2014-12-12 (inclusive). Note that only the first few columns are shown due to the limitation of page width. * Note that the right of the table has been cropped to fit on the page. *

Because the table is in date order, we can be confident that only the rows with dates between 8 December 2014 and 12 December 2014 (inclusive) will be returned. However if the table had not been in date order, we would have needed to sort it first, like this:

london = london.sort_index()

Now there is a datetime64 index, let’s plot ’ Max Wind SpeedKm/h ’again:

In []:

london['Max Wind SpeedKm/h'].plot(grid=True, figsize=(10,5))

Out[]:

Figure 14

Chart of the values in the Max Wind SpeedKm/h column of the london dataframe. Note that the legend for the x-axis has changed from numbers to month names. Now it is much clearer that the worst winds were in mid-February.

## Exercise 6 Changing a dataframe’s index¶

### Question¶

Now try Exercise 6 in the Exercise notebook 2.

#### 2.2 The project¶

Your project this week is to find out what would have been the best two weeks of weather for a 2014 vacation in a capital of a BRICS country.

Figure 15

An image of filter like diagonal strips across various skies such as an orange sunset, a storm and a clear blue sky I’ve written up my analysis of the best two weeks of weather in London, UK, which you can open in project: 2: Holiday weather.

The structure is very simple: besides the introduction and the conclusions, there is one section for each step of the analysis – obtaining, cleaning and visualising the data.

Once you’ve worked through my analysis you should open a dataset for just one of the BRICS capitals: Brasilia, Moscow, Delhi, Beijing or Cape Town. The choice of capital is up to you. You should then work out the best two weeks, according to the weather, to choose for a two-week holiday in your chosen capital city.

• Right click on the name of your chosen capital city above