Scientific computing with Python - Introduction to Pandas
Materials by Mike Hansen and Aleksandra Pawlik
There are various very useful libraries written in Python that are widely used for scientific computing. If you are unfamiliar with the concept of a "library" in programming, you can think of it as a collection of modules with functions that do specific things (for example, statistical tests) that you can call from the code you write yourself.
For example, when dealing with numeric matrices and vectors in Python, NumPy makes life a lot easier. For more complex data, however, it leaves a lot to be desired. If you're used to working with data frames in R, doing data analysis directly with NumPy feels like a step back.
Fortunately, some nice folks have written the Python Data Analysis Library (a.k.a. pandas).
Pandas provides an R-like DataFrame, produces high quality plots with matplotlib, and integrates nicely with other libraries that expect NumPy arrays.
In this tutorial, we'll go through the basics of pandas using a year's worth of weather data from Weather Underground. Pandas has a lot of functionality, so we'll only be able to cover a small fraction of what you can do. Check out the (very readable) pandas docs if you want to learn more.
The IPython Notebook file and the sample data files for this tutorial can be found in the course materials:
2014-05-08-soton/lessons/nocs/pandas/
Getting Started
OK, let's get started by importing the pandas library
import pandas
Importing a library is like getting a piece of lab equipment out of a storage locker and setting it up on the bench. Once it's done, we can ask the library to read our data file for us.
Let's read in our data then.
Because it's in a CSV file, we can use pandas' read_csv function to pull it directly into a DataFrame.
data = pandas.read_csv("data/weather_year.csv")
We can get a summary of the DataFrame by printing the object.
data
Output:
<class 'pandas.core.frame.DataFrame'>
Int64Index: 366 entries, 0 to 365
Data columns:
EDT 366 non-null values
Max TemperatureF 366 non-null values
Mean TemperatureF 366 non-null values
Min TemperatureF 366 non-null values
Max Dew PointF 366 non-null values
MeanDew PointF 366 non-null values
Min DewpointF 366 non-null values
Max Humidity 366 non-null values
Mean Humidity 366 non-null values
Min Humidity 366 non-null values
Max Sea Level PressureIn 366 non-null values
Mean Sea Level PressureIn 366 non-null values
Min Sea Level PressureIn 366 non-null values
Max VisibilityMiles 366 non-null values
Mean VisibilityMiles 366 non-null values
Min VisibilityMiles 366 non-null values
Max Wind SpeedMPH 366 non-null values
Mean Wind SpeedMPH 366 non-null values
Max Gust SpeedMPH 365 non-null values
PrecipitationIn 366 non-null values
CloudCover 366 non-null values
Events 162 non-null values
WindDirDegrees 366 non-null values
dtypes: float64(4), int64(16), object(3)
This gives us a lot of information. First, we can see that there are 366 rows (entries) – a year and a day's worth of weather. Each column is printed along with however many "non-null" values are present.
We'll talk more about null (or missing) values in pandas later, but for now we can note that only the "Max Gust SpeedMPH" and "Events" columns have fewer than 366 non-null values.
Lastly, the data types (dtypes) of the columns are printed at the very bottom. We can see that there are 4 float64, 16 int64, and 3 object columns.
len(data)
Output:
366
Using len on a DataFrame will give you the number of rows. You can get the column names using the columns property.
data.columns
Output:
Index([EDT, Max TemperatureF, Mean TemperatureF, Min TemperatureF,
Max Dew PointF, MeanDew PointF, Min DewpointF, Max Humidity,
Mean Humidity, Min Humidity, Max Sea Level PressureIn,
Mean Sea Level PressureIn, Min Sea Level PressureIn,
Max VisibilityMiles, Mean VisibilityMiles, Min VisibilityMiles,
Max Wind SpeedMPH, Mean Wind SpeedMPH, Max Gust SpeedMPH,
PrecipitationIn, CloudCover, Events, WindDirDegrees], dtype=object)
Columns can be accessed in two ways. The first is using the DataFrame like a dictionary with string keys:
data["EDT"]
Output:
0 2012-3-10
1 2012-3-11
2 2012-3-12
3 2012-3-13
4 2012-3-14
5 2012-3-15
6 2012-3-16
7 2012-3-17
8 2012-3-18
9 2012-3-19
10 2012-3-20
11 2012-3-21
12 2012-3-22
13 2012-3-23
14 2012-3-24
...
351 2013-2-24
352 2013-2-25
353 2013-2-26
354 2013-2-27
355 2013-2-28
356 2013-3-1
357 2013-3-2
358 2013-3-3
359 2013-3-4
360 2013-3-5
361 2013-3-6
362 2013-3-7
363 2013-3-8
364 2013-3-9
365 2013-3-10
Name: EDT, Length: 366
You can get multiple columns out at the same time by passing in a list of strings.
data[["EDT", "Mean TemperatureF"]]
Output:
<class 'pandas.core.frame.DataFrame'>
Int64Index: 366 entries, 0 to 365
Data columns:
EDT 366 non-null values
Mean TemperatureF 366 non-null values
dtypes: int64(1), object(1)
The second way to access columns is using the dot syntax. This only works if your column name could also be a Python variable name (i.e., no spaces), and if it doesn't collide with another DataFrame property or function name (e.g., count, sum).
data.EDT
Output:
0 2012-3-10
1 2012-3-11
2 2012-3-12
3 2012-3-13
4 2012-3-14
5 2012-3-15
6 2012-3-16
7 2012-3-17
8 2012-3-18
9 2012-3-19
10 2012-3-20
11 2012-3-21
12 2012-3-22
13 2012-3-23
14 2012-3-24
...
351 2013-2-24
352 2013-2-25
353 2013-2-26
354 2013-2-27
355 2013-2-28
356 2013-3-1
357 2013-3-2
358 2013-3-3
359 2013-3-4
360 2013-3-5
361 2013-3-6
362 2013-3-7
363 2013-3-8
364 2013-3-9
365 2013-3-10
Name: EDT, Length: 366
We'll be mostly using the dot syntax here because you can auto-complete the names in IPython and IPython Notebook. The first pandas function we'll learn about is head(). This gives us the first 5 items in a column (or the first 5 rows in the DataFrame).
data.EDT.head()
Output:
0 2012-3-10
1 2012-3-11
2 2012-3-12
3 2012-3-13
4 2012-3-14
Name: EDT
Passing in a number n gives us the first n items in the column. There is also a corresponding tail() method that gives the last n items or rows.
data.EDT.head(10)
Output:
0 2012-3-10
1 2012-3-11
2 2012-3-12
3 2012-3-13
4 2012-3-14
5 2012-3-15
6 2012-3-16
7 2012-3-17
8 2012-3-18
9 2012-3-19
Name: EDT
This also works with the dictionary syntax.
data["Mean TemperatureF"].head()
Output:
0 40
1 49
2 62
3 63
4 62
Name: Mean TemperatureF
Exercise 1:
How would we get the second to last date (EDT) in the dataset?
# Combine head() and tail()
last_two_dates = data.EDT.tail(2)
second_to_last_date = last_two_dates.head(1)
print second_to_last_date
Output:
364 2013-3-9
Name: EDT
If the data in the column is numeric, you can use describe() to get some stats on it.
data["Mean TemperatureF"].describe()
Output:
count 366.000000
mean 55.683060
std 18.436506
min 11.000000
25% 41.000000
50% 59.000000
75% 70.750000
max 89.000000
Fun with Columns
The column names in data are a little unweildy, so we're going to rename them. You can rename individual columns with the rename() method of the DataFrame.
data = data.rename(columns={ "Max TemperatureF": "max_temp", "Min TemperatureF": "min_temp" })
data.columns
Output:
Index([EDT, max_temp, Mean TemperatureF, min_temp, Max Dew PointF,
MeanDew PointF, Min DewpointF, Max Humidity, Mean Humidity,
Min Humidity, Max Sea Level PressureIn,
Mean Sea Level PressureIn, Min Sea Level PressureIn,
Max VisibilityMiles, Mean VisibilityMiles, Min VisibilityMiles,
Max Wind SpeedMPH, Mean Wind SpeedMPH, Max Gust SpeedMPH,
PrecipitationIn, CloudCover, Events, WindDirDegrees], dtype=object)
You can see that the max and mean temperature columns are now called max_temp and mean_temp respectively.
Notice that we assigned the results of data.rename(...) to data. Most DataFrame methods do not modify the DataFrame directly, and instead return a copy or a filtered view.
Let's change all of our column names at the same time. This is as easy as assigning a new list of column names to the columns property of the DataFrame.
data.columns = ["date", "max_temp", "mean_temp", "min_temp", "max_dew",
"mean_dew", "min_dew", "max_humidity", "mean_humidity",
"min_humidity", "max_pressure", "mean_pressure",
"min_pressure", "max_visibilty", "mean_visibility",
"min_visibility", "max_wind", "mean_wind", "min_wind",
"precipitation", "cloud_cover", "events", "wind_dir"]
These should be in the same order as the original columns. Let's take another look at our DataFrame summary.
data
Output:
<class 'pandas.core.frame.DataFrame'>
Int64Index: 366 entries, 0 to 365
Data columns:
date 366 non-null values
max_temp 366 non-null values
mean_temp 366 non-null values
min_temp 366 non-null values
max_dew 366 non-null values
mean_dew 366 non-null values
min_dew 366 non-null values
max_humidity 366 non-null values
mean_humidity 366 non-null values
min_humidity 366 non-null values
max_pressure 366 non-null values
mean_pressure 366 non-null values
min_pressure 366 non-null values
max_visibilty 366 non-null values
mean_visibility 366 non-null values
min_visibility 366 non-null values
max_wind 366 non-null values
mean_wind 366 non-null values
min_wind 365 non-null values
precipitation 366 non-null values
cloud_cover 366 non-null values
events 162 non-null values
wind_dir 366 non-null values
dtypes: float64(4), int64(16), object(3)
Now our columns can all be accessed using the dot syntax!
data.mean_temp.head()
Output:
0 40
1 49
2 62
3 63
4 62
Name: mean_temp
There are lots useful methods on columns, such as std() to get the standard deviation. Most of pandas' methods will happily ignore missing values like NaN.
data.mean_temp.std()
Output:
18.436505996251075
Some methods, like plot() and hist() produce plots using matplotlib. We'll go over plotting in more detail later.
data.mean_temp.hist()
Output:
<matplotlib.axes.AxesSubplot at 0x3f8a950>
By the way, many of the column-specific methods also work on the entire DataFrame. Instead of a single number, you'll get a result for each column.
data.std()
Output:
max_temp 20.361247
mean_temp 18.436506
min_temp 17.301141
max_dew 16.397178
mean_dew 16.829996
min_dew 17.479449
max_humidity 9.108438
mean_humidity 9.945591
min_humidity 15.360261
max_pressure 0.172189
mean_pressure 0.174112
min_pressure 0.182476
max_visibilty 0.073821
mean_visibility 1.875406
min_visibility 3.792219
max_wind 5.564329
mean_wind 3.200940
min_wind 8.131092
cloud_cover 2.707261
wind_dir 94.045080
If you only want to compute something over couple of the columns, use the dictionary syntax:
data[["max_temp", "max_pressure", "max_wind"]].std()
Output:
max_temp 20.361247
max_pressure 0.172189
max_wind 5.564329
Exercise 2:
What is the range of temperatures in the dataset?
Hint: columns have max() and min() methods.
hottest_temp = data.max_temp.max() # Highest of the highs
coldest_temp = data.min_temp.min() # Lowest of the lows
print "Temperature range:", hottest_temp - coldest_temp, "degrees F"
Output:
Temperature range: 105 degrees F
Bulk Operations with apply()
Built-on methods like sum() and std() work on entire columns. We can run our own functions across all values in a column (or row) using apply().
To give you an idea of how this works, let's consider the "date" column in our DataFrame (formally "EDT").
data.date.head()
Output:
0 2012-3-10
1 2012-3-11
2 2012-3-12
3 2012-3-13
4 2012-3-14
Name: date
We can use the values property of the column to get a list of values for the column. Inspecting the first value reveals that these are strings with a particular format.
first_date = data.date.values[0]
print first_date, "is a", type(first_date)
Output:
2012-3-10 is a <type 'str'>
The strptime function from the datetime module will make quick work of this date string. There are many more shortcuts available for strptime.
# Import the datetime class from the datetime module
from datetime import datetime
# Convert date string to datetime object
datetime.strptime(first_date, "%Y-%m-%d")
Output:
datetime.datetime(2012, 3, 10, 0, 0)
Using the apply() method, which takes a function (without the parentheses), we can apply strptime to each value in the column. We'll overwrite the string date values with their Python datetime equivalents.
# Define a function to convert strings to dates
def string_to_date(date_string):
return datetime.strptime(date_string, "%Y-%m-%d")
# Run the function on every date string and overwrite the column
data.date = data.date.apply(string_to_date)
sdata.date.head()
Output:
0 2012-03-10 00:00:00
1 2012-03-11 00:00:00
2 2012-03-12 00:00:00
3 2012-03-13 00:00:00
4 2012-03-14 00:00:00
Name: date
Let's go one step futher. Each row in our DateFrame represents the weather from a single day. Each row in a DataFrame is associated with an index, which is a label that uniquely identifies a row.
Our row indices up to now have been auto-generated by pandas, and are simply integers from 0 to 365. If we use dates instead of integers for our index, we will get some extra benefits from pandas when plotting later on. Overwriting the index is as easy as assigning to the index property of the DataFrame.
data.index = data.date
data.index
Output:
<class 'pandas.tseries.index.DatetimeIndex'>
[2012-03-10 00:00:00, ..., 2013-03-10 00:00:00]
Length: 366, Freq: None, Timezone: None
Now we can quickly look up a row by its date with the ix[] property.
data.ix[datetime(2012, 8, 19)]
We can even slice out a while month using a list-like syntax:
data[datetime(2012, 4, 1):datetime(2012, 4, 30)]
Output:
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 30 entries, 2012-04-01 00:00:00 to 2012-04-30 00:00:00
Data columns:
date 30 non-null values
max_temp 30 non-null values
mean_temp 30 non-null values
min_temp 30 non-null values
max_dew 30 non-null values
mean_dew 30 non-null values
min_dew 30 non-null values
max_humidity 30 non-null values
mean_humidity 30 non-null values
min_humidity 30 non-null values
max_pressure 30 non-null values
mean_pressure 30 non-null values
min_pressure 30 non-null values
max_visibilty 30 non-null values
mean_visibility 30 non-null values
min_visibility 30 non-null values
max_wind 30 non-null values
mean_wind 30 non-null values
min_wind 30 non-null values
precipitation 30 non-null values
cloud_cover 30 non-null values
events 14 non-null values
wind_dir 30 non-null values
dtypes: float64(4), int64(16), object(3)
With all of the dates in the index now, we no longer need the "date" column. Let's drop it.
data = data.drop("date", axis=1)
data.columns
Output:
Index([max_temp, mean_temp, min_temp, max_dew, mean_dew, min_dew,
max_humidity, mean_humidity, min_humidity, max_pressure,
mean_pressure, min_pressure, max_visibilty, mean_visibility,
min_visibility, max_wind, mean_wind, min_wind, precipitation,
cloud_cover, events, wind_dir], dtype=object)
Note that we need to pass in axis=1 in order to drop a column. For more details, check out the documentation for drop.
pandas.DataFrame.drop?
Exercise 3:
Print out the cloud cover for each day in May.
Hint: you can make datetime objects with the datetime(year, month, day) function
datetime(2012, 5, 1) # May 1st of 2012
data[datetime(2012, 5, 1):datetime(2012, 5, 31)].cloud_cover
Output:
date
2012-05-01 6
2012-05-02 1
2012-05-03 0
2012-05-04 6
2012-05-05 3
2012-05-06 0
2012-05-07 5
2012-05-08 4
2012-05-09 3
2012-05-10 1
2012-05-11 0
2012-05-12 1
2012-05-13 4
2012-05-14 4
2012-05-15 0
2012-05-16 0
2012-05-17 0
2012-05-18 0
2012-05-19 0
2012-05-20 1
2012-05-21 4
2012-05-22 2
2012-05-23 0
2012-05-24 0
2012-05-25 2
2012-05-26 0
2012-05-27 0
2012-05-28 0
2012-05-29 4
2012-05-30 1
2012-05-31 4
Name: cloud_cover
Handling Missing Values
Pandas considers values like NaN and None to represent missing data. The pandas.isnull function can be used to tell whether or not a value is missing.
Let's use apply() across all of the columns in our DataFrame to figure out which values are missing.
empty = data.apply(lambda col: pandas.isnull(col))
empty
Output:
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 366 entries, 2012-03-10 00:00:00 to 2013-03-10 00:00:00
Freq: D
Data columns:
max_temp 366 non-null values
mean_temp 366 non-null values
min_temp 366 non-null values
max_dew 366 non-null values
mean_dew 366 non-null values
min_dew 366 non-null values
max_humidity 366 non-null values
mean_humidity 366 non-null values
min_humidity 366 non-null values
max_pressure 366 non-null values
mean_pressure 366 non-null values
min_pressure 366 non-null values
max_visibilty 366 non-null values
mean_visibility 366 non-null values
min_visibility 366 non-null values
max_wind 366 non-null values
mean_wind 366 non-null values
min_wind 366 non-null values
precipitation 366 non-null values
cloud_cover 366 non-null values
events 366 non-null values
wind_dir 366 non-null values
dtypes: bool(22)
We got back a dataframe (empty) with boolean values for all 22 columns and 366 rows. Inspecting the first 10 values of the "events", column we can see that there are some missing values because a True was returned from pandas.isnull.
empty.events.head(10)
Output:
date
2012-03-10 True
2012-03-11 False
2012-03-12 False
2012-03-13 True
2012-03-14 True
2012-03-15 False
2012-03-16 True
2012-03-17 False
2012-03-18 False
2012-03-19 True
Freq: D, Name: events
Looking at the corresponding rows in the original DataFrame reveals that pandas has filled in NaN for empty values in the "events" column.
data.events.head(10)
Output:
date
2012-03-10 NaN
2012-03-11 Rain
2012-03-12 Rain
2012-03-13 NaN
2012-03-14 NaN
2012-03-15 Rain-Thunderstorm
2012-03-16 NaN
2012-03-17 Fog-Thunderstorm
2012-03-18 Rain
2012-03-19 NaN
Freq: D, Name: events
This isn't exactly what we want. One option is to drop all rows in the DataFrame with missing "events" values.
data.dropna(subset=["events"])
Output:
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 162 entries, 2012-03-11 00:00:00 to 2013-03-06 00:00:00
Data columns:
max_temp 162 non-null values
mean_temp 162 non-null values
min_temp 162 non-null values
max_dew 162 non-null values
mean_dew 162 non-null values
min_dew 162 non-null values
max_humidity 162 non-null values
mean_humidity 162 non-null values
min_humidity 162 non-null values
max_pressure 162 non-null values
mean_pressure 162 non-null values
min_pressure 162 non-null values
max_visibilty 162 non-null values
mean_visibility 162 non-null values
min_visibility 162 non-null values
max_wind 162 non-null values
mean_wind 162 non-null values
min_wind 162 non-null values
precipitation 162 non-null values
cloud_cover 162 non-null values
events 162 non-null values
wind_dir 162 non-null values
dtypes: float64(4), int64(16), object(2)
The DataFrame we get back has only 162 rows, so we can infer that there were 366 - 162 = 204 missing values in the "events" column. Note that this didn't affect data; we're just looking at a copy.
Instead of dropping the rows with missing values, let's fill them with empty strings (you'll see why in a moment). This is easily done with the fillna() function. We'll go ahead and overwrite the "events" column with empty string missing values instead of NaN.
data.events = data.events.fillna("")
data.events.head(10)
Output:
date
2012-03-10
2012-03-11 Rain
2012-03-12 Rain
2012-03-13
2012-03-14
2012-03-15 Rain-Thunderstorm
2012-03-16
2012-03-17 Fog-Thunderstorm
2012-03-18 Rain
2012-03-19
Freq: D, Name: events
Accessing Individual Rows
Sometimes you need to access individual rows in your DataFrame. The irow() function lets you grab the ith row from a DataFrame (starting from 0).
data.irow(0)
Output:
max_temp 56
mean_temp 40
min_temp 24
max_dew 24
mean_dew 20
min_dew 16
max_humidity 74
mean_humidity 50
min_humidity 26
max_pressure 30.53
mean_pressure 30.45
min_pressure 30.34
max_visibilty 10
mean_visibility 10
min_visibility 10
max_wind 13
mean_wind 6
min_wind 17
precipitation 0.00
cloud_cover 0
events
wind_dir 138
Name: 2012-03-10 00:00:00
Of course, another option is to use the index.
data.ix[datetime(2013, 1, 1)]
Output:
max_temp 32
mean_temp 26
min_temp 20
max_dew 31
mean_dew 25
min_dew 16
max_humidity 92
mean_humidity 83
min_humidity 74
max_pressure 30.2
mean_pressure 30.11
min_pressure 30.04
max_visibilty 9
mean_visibility 5
min_visibility 2
max_wind 14
mean_wind 5
min_wind 15
precipitation T
cloud_cover 8
events
wind_dir 353
Name: 2013-01-01 00:00:00
You can iterate over each row in the DataFrame with iterrows(). Note that this function returns both the index and the row. Also, you must access columns in the row you get back from iterrows() with the dictionary syntax.
num_rain = 0
for idx, row in data.iterrows():
if "Rain" in row["events"]:
num_rain += 1
print "Days with rain:", num_rain
Output:
Days with rain: 121
Exercise 4:
Was there any November rain?
Hint: check out the strftime() function on datetime objects and the documentation
d = datetime(2012, 1, 1)
d.strftime("%B")
november_rain = False
for date_idx, row in data.iterrows():
if date_idx.strftime("%B") == "November" and "Rain" in row["events"]:
november_rain = True
if november_rain:
print "There was rain in November"
else:
print "There was *not* rain in November"
Output:
There was rain in November
Filtering
Most of your time using pandas will likely be devoted to selecting rows of interest from a DataFrame. In addition to strings, the dictionary syntax accepts things like this:
freezing_days = data[data.max_temp <= 32]
freezing_days
Output:
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 21 entries, 2012-11-24 00:00:00 to 2013-03-06 00:00:00
Data columns:
max_temp 21 non-null values
mean_temp 21 non-null values
min_temp 21 non-null values
max_dew 21 non-null values
mean_dew 21 non-null values
min_dew 21 non-null values
max_humidity 21 non-null values
mean_humidity 21 non-null values
min_humidity 21 non-null values
max_pressure 21 non-null values
mean_pressure 21 non-null values
min_pressure 21 non-null values
max_visibilty 21 non-null values
mean_visibility 21 non-null values
min_visibility 21 non-null values
max_wind 21 non-null values
mean_wind 21 non-null values
min_wind 21 non-null values
precipitation 21 non-null values
cloud_cover 21 non-null values
events 21 non-null values
wind_dir 21 non-null values
dtypes: float64(4), int64(16), object(2)
We get back another DataFrame with fewer rows (21 in this case). This DataFrame can be filtered down even more.
freezing_days[freezing_days.min_temp >= 20]
Output:
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 7 entries, 2012-11-24 00:00:00 to 2013-03-06 00:00:00
Data columns:
max_temp 7 non-null values
mean_temp 7 non-null values
min_temp 7 non-null values
max_dew 7 non-null values
mean_dew 7 non-null values
min_dew 7 non-null values
max_humidity 7 non-null values
mean_humidity 7 non-null values
min_humidity 7 non-null values
max_pressure 7 non-null values
mean_pressure 7 non-null values
min_pressure 7 non-null values
max_visibilty 7 non-null values
mean_visibility 7 non-null values
min_visibility 7 non-null values
max_wind 7 non-null values
mean_wind 7 non-null values
min_wind 7 non-null values
precipitation 7 non-null values
cloud_cover 7 non-null values
events 7 non-null values
wind_dir 7 non-null values
dtypes: float64(4), int64(16), object(2)
Or, using boolean operations, we could apply both filters to the original DataFrame at the same time.
data[(data.max_temp <= 32) & (data.min_temp >= 20)]
Output:
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 7 entries, 2012-11-24 00:00:00 to 2013-03-06 00:00:00
Data columns:
max_temp 7 non-null values
mean_temp 7 non-null values
min_temp 7 non-null values
max_dew 7 non-null values
mean_dew 7 non-null values
min_dew 7 non-null values
max_humidity 7 non-null values
mean_humidity 7 non-null values
min_humidity 7 non-null values
max_pressure 7 non-null values
mean_pressure 7 non-null values
min_pressure 7 non-null values
max_visibilty 7 non-null values
mean_visibility 7 non-null values
min_visibility 7 non-null values
max_wind 7 non-null values
mean_wind 7 non-null values
min_wind 7 non-null values
precipitation 7 non-null values
cloud_cover 7 non-null values
events 7 non-null values
wind_dir 7 non-null values
dtypes: float64(4), int64(16), object(2)
It's important to understand what's really going on underneath with filtering. Let's look at what kind of object we actually get back when creating a filter.
temp_max = data.max_temp <= 32
type(temp_max)
Output:
pandas.core.series.TimeSeries
This is a pandas Series object, which is the one-dimensional equivalent of a DataFrame. Because our DataFrame uses datetime objects for the index, we have a specialized TimeSeries object.
What's inside the filter?
temp_max
Output:
date
2012-03-10 False
2012-03-11 False
2012-03-12 False
2012-03-13 False
2012-03-14 False
2012-03-15 False
2012-03-16 False
2012-03-17 False
2012-03-18 False
2012-03-19 False
2012-03-20 False
2012-03-21 False
2012-03-22 False
2012-03-23 False
2012-03-24 False
...
2013-02-24 False
2013-02-25 False
2013-02-26 False
2013-02-27 False
2013-02-28 False
2013-03-01 False
2013-03-02 True
2013-03-03 False
2013-03-04 False
2013-03-05 False
2013-03-06 True
2013-03-07 False
2013-03-08 False
2013-03-09 False
2013-03-10 False
Freq: D, Name: max_temp, Length: 366
Our filter is nothing more than a Series with a boolean value for every item in the index. When we "run the filter" as so:
data[temp_max]
Output:
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 21 entries, 2012-11-24 00:00:00 to 2013-03-06 00:00:00
Data columns:
max_temp 21 non-null values
mean_temp 21 non-null values
min_temp 21 non-null values
max_dew 21 non-null values
mean_dew 21 non-null values
min_dew 21 non-null values
max_humidity 21 non-null values
mean_humidity 21 non-null values
min_humidity 21 non-null values
max_pressure 21 non-null values
mean_pressure 21 non-null values
min_pressure 21 non-null values
max_visibilty 21 non-null values
mean_visibility 21 non-null values
min_visibility 21 non-null values
max_wind 21 non-null values
mean_wind 21 non-null values
min_wind 21 non-null values
precipitation 21 non-null values
cloud_cover 21 non-null values
events 21 non-null values
wind_dir 21 non-null values
dtypes: float64(4), int64(16), object(2)
pandas lines up the rows of the DataFrame and the filter using the index, and then keeps the rows with a True filter value. That's it.
Let's create another filter.
temp_min = data.min_temp >= 20
temp_min
Output:
date
2012-03-10 True
2012-03-11 True
2012-03-12 True
2012-03-13 True
2012-03-14 True
2012-03-15 True
2012-03-16 True
2012-03-17 True
2012-03-18 True
2012-03-19 True
2012-03-20 True
2012-03-21 True
2012-03-22 True
2012-03-23 True
2012-03-24 True
...
2013-02-24 True
2013-02-25 True
2013-02-26 True
2013-02-27 True
2013-02-28 True
2013-03-01 True
2013-03-02 True
2013-03-03 False
2013-03-04 False
2013-03-05 True
2013-03-06 True
2013-03-07 True
2013-03-08 True
2013-03-09 True
2013-03-10 True
Freq: D, Name: min_temp, Length: 366
Now we can see what the boolean operations are doing. Something like & (not and)…
temp_min & temp_max
Output:
date
2012-03-10 False
2012-03-11 False
2012-03-12 False
2012-03-13 False
2012-03-14 False
2012-03-15 False
2012-03-16 False
2012-03-17 False
2012-03-18 False
2012-03-19 False
2012-03-20 False
2012-03-21 False
2012-03-22 False
2012-03-23 False
2012-03-24 False
...
2013-02-24 False
2013-02-25 False
2013-02-26 False
2013-02-27 False
2013-02-28 False
2013-03-01 False
2013-03-02 True
2013-03-03 False
2013-03-04 False
2013-03-05 False
2013-03-06 True
2013-03-07 False
2013-03-08 False
2013-03-09 False
2013-03-10 False
Freq: D, Length: 366
…is just lining up the two filters using the index, performing a boolean AND operation, and returning the result as another Series.
We can do other boolean operations too, like OR:
temp_min | temp_max
Output:
date
2012-03-10 True
2012-03-11 True
2012-03-12 True
2012-03-13 True
2012-03-14 True
2012-03-15 True
2012-03-16 True
2012-03-17 True
2012-03-18 True
2012-03-19 True
2012-03-20 True
2012-03-21 True
2012-03-22 True
2012-03-23 True
2012-03-24 True
...
2013-02-24 True
2013-02-25 True
2013-02-26 True
2013-02-27 True
2013-02-28 True
2013-03-01 True
2013-03-02 True
2013-03-03 False
2013-03-04 False
2013-03-05 True
2013-03-06 True
2013-03-07 True
2013-03-08 True
2013-03-09 True
2013-03-10 True
Freq: D, Length: 366
Because the result is just another Series, we have all of the regular pandas functions at our disposal. The any() function returns True if any value in the Series is True.
temp_both = temp_min & temp_max
temp_both.any()
Output:
True
Sometimes filters aren't so intuitive. This (sadly) doesn't work:
try:
data["Rain" in data.events]
except:
pass # "KeyError: no item named False"
We can wrap it up in an apply() call fairly easily, though:
def has_rain(event_string):
return "Rain" in event_string
data[data.events.apply(has_rain)]
Output:
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 121 entries, 2012-03-11 00:00:00 to 2013-03-05 00:00:00
Data columns:
max_temp 121 non-null values
mean_temp 121 non-null values
min_temp 121 non-null values
max_dew 121 non-null values
mean_dew 121 non-null values
min_dew 121 non-null values
max_humidity 121 non-null values
mean_humidity 121 non-null values
min_humidity 121 non-null values
max_pressure 121 non-null values
mean_pressure 121 non-null values
min_pressure 121 non-null values
max_visibilty 121 non-null values
mean_visibility 121 non-null values
min_visibility 121 non-null values
max_wind 121 non-null values
mean_wind 121 non-null values
min_wind 121 non-null values
precipitation 121 non-null values
cloud_cover 121 non-null values
events 121 non-null values
wind_dir 121 non-null values
dtypes: float64(4), int64(16), object(2)
Exercise 5:
Before starting the exercise, let's convert the precipitation column in the dataset to floating point numbers. It's currently full of strings because of the "T" value, which stands for "trace amount of precipitation."
data.precipitation.head()
Output:
date
2012-03-10 0.00
2012-03-11 T
2012-03-12 0.03
2012-03-13 0.00
2012-03-14 0.00
Freq: D, Name: precipitation
We'll replace "T" with a very small number, and convert the rest of the strings to floats:
# Convert precipitation to floating point number
# "T" means "trace of precipitation"
def precipitation_to_float(precip_str):
if precip_str == "T":
return 1e-10 # Very small value
return float(precip_str)
data.precipitation = data.precipitation.apply(precipitation_to_float)
data.precipitation.head()
Output:
date
2012-03-10 0.000000e+00
2012-03-11 1.000000e-10
2012-03-12 3.000000e-02
2012-03-13 0.000000e+00
2012-03-14 0.000000e+00
Freq: D, Name: precipitation
Now on to the exercise.
What was the coldest it ever got when there was no cloud cover and no precipitation?
no_cloud_cover = (data.cloud_cover == 0)
no_precipitation = (data.precipitation == 0)
coldest_temp = data[no_cloud_cover & no_precipitation].min_temp.min()
print "Coldest without cloud cover and precipitation was", coldest_temp, "degree(s) F"
Output:
Coldest without cloud cover and precipitation was 1 degree(s) F
Grouping
Besides apply(), another great DataFrame function is groupby().
It will group a DataFrame by one or more columns, and let you iterate through each group.
As an example, let's group our DataFrame by the "cloud_cover" column (a value ranging from 0 to 8).
cover_temps = {}
for cover, cover_data in data.groupby("cloud_cover"):
cover_temps[cover] = cover_data.mean_temp.mean() # The mean mean temp!
cover_temps
Output:
{0: 59.730769230769234,
1: 61.415094339622641,
2: 59.727272727272727,
3: 58.0625,
4: 51.5,
5: 50.827586206896555,
6: 57.727272727272727,
7: 46.5,
8: 40.909090909090907}
When you iterate through the result of groupby(), you will get a tuple.
The first item is the column value, and the second item is a filtered DataFrame (where the column equals the first tuple value).
You can group by more than one column as well.
In this case, the first tuple item returned by groupby() will itself be a tuple with the value of each column.
for (cover, events), group_data in data.groupby(["cloud_cover", "events"]):
print "Cover: {0}, Events: {1}, Count: {2}".format(cover, events, len(group_data))
Output:
Cover: 0, Events: , Count: 99
Cover: 0, Events: Fog, Count: 2
Cover: 0, Events: Rain, Count: 2
Cover: 0, Events: Thunderstorm, Count: 1
Cover: 1, Events: , Count: 35
Cover: 1, Events: Fog, Count: 5
Cover: 1, Events: Fog-Rain, Count: 1
Cover: 1, Events: Rain, Count: 4
Cover: 1, Events: Rain-Thunderstorm, Count: 2
Cover: 1, Events: Thunderstorm, Count: 6
Cover: 2, Events: , Count: 20
Cover: 2, Events: Fog, Count: 1
Cover: 2, Events: Rain, Count: 5
Cover: 2, Events: Rain-Thunderstorm, Count: 4
Cover: 2, Events: Snow, Count: 1
Cover: 2, Events: Thunderstorm, Count: 2
Cover: 3, Events: , Count: 12
Cover: 3, Events: Fog, Count: 2
Cover: 3, Events: Fog-Rain-Thunderstorm, Count: 3
Cover: 3, Events: Fog-Thunderstorm, Count: 1
Cover: 3, Events: Rain, Count: 9
Cover: 3, Events: Rain-Thunderstorm, Count: 4
Cover: 3, Events: Snow, Count: 1
Cover: 4, Events: , Count: 16
Cover: 4, Events: Fog, Count: 3
Cover: 4, Events: Fog-Rain, Count: 2
Cover: 4, Events: Fog-Rain-Thunderstorm, Count: 2
Cover: 4, Events: Rain, Count: 10
Cover: 4, Events: Rain-Thunderstorm, Count: 6
Cover: 4, Events: Snow, Count: 1
Cover: 5, Events: , Count: 9
Cover: 5, Events: Fog-Rain, Count: 1
Cover: 5, Events: Fog-Rain-Snow, Count: 1
Cover: 5, Events: Rain, Count: 13
Cover: 5, Events: Rain-Thunderstorm, Count: 3
Cover: 5, Events: Snow, Count: 2
Cover: 6, Events: , Count: 3
Cover: 6, Events: Fog-Rain, Count: 2
Cover: 6, Events: Fog-Rain-Snow, Count: 1
Cover: 6, Events: Fog-Rain-Thunderstorm, Count: 2
Cover: 6, Events: Rain, Count: 9
Cover: 6, Events: Rain-Thunderstorm, Count: 4
Cover: 6, Events: Snow, Count: 1
Cover: 7, Events: , Count: 5
Cover: 7, Events: Fog-Rain, Count: 1
Cover: 7, Events: Fog-Rain-Thunderstorm, Count: 1
Cover: 7, Events: Fog-Snow, Count: 3
Cover: 7, Events: Rain, Count: 6
Cover: 7, Events: Rain-Thunderstorm, Count: 3
Cover: 7, Events: Snow, Count: 1
Cover: 8, Events: , Count: 5
Cover: 8, Events: Fog-Rain, Count: 4
Cover: 8, Events: Fog-Rain-Snow, Count: 1
Cover: 8, Events: Fog-Rain-Snow-Thunderstorm, Count: 1
Cover: 8, Events: Fog-Snow, Count: 2
Cover: 8, Events: Rain, Count: 11
Cover: 8, Events: Rain-Snow, Count: 3
Cover: 8, Events: Snow, Count: 6
Creating New Columns
Weather events in our DataFrame are stored in strings like "Rain-Thunderstorm" to represent that it rained and there was a thunderstorm that day. Let's split them out into boolean "rain", "thunderstorm", etc. columns.
First, let's discover the different kinds of weather events we have with unique().
data.events.unique()
Output:
array([, Rain, Rain-Thunderstorm, Fog-Thunderstorm, Fog-Rain, Thunderstorm,
Fog-Rain-Thunderstorm, Fog, Fog-Rain-Snow,
Fog-Rain-Snow-Thunderstorm, Fog-Snow, Snow, Rain-Snow], dtype=object)
Looks like we have "Rain", "Thunderstorm", "Fog", and "Snow" events. Creating a new column for each of these event kinds is a piece of cake with the dictionary syntax.
for event_kind in ["Rain", "Thunderstorm", "Fog", "Snow"]:
col_name = event_kind.lower() # Turn "Rain" into "rain", etc.
data[col_name] = data.events.apply(lambda e: event_kind in e)
data
Output:
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 366 entries, 2012-03-10 00:00:00 to 2013-03-10 00:00:00
Freq: D
Data columns:
max_temp 366 non-null values
mean_temp 366 non-null values
min_temp 366 non-null values
max_dew 366 non-null values
mean_dew 366 non-null values
min_dew 366 non-null values
max_humidity 366 non-null values
mean_humidity 366 non-null values
min_humidity 366 non-null values
max_pressure 366 non-null values
mean_pressure 366 non-null values
min_pressure 366 non-null values
max_visibilty 366 non-null values
mean_visibility 366 non-null values
min_visibility 366 non-null values
max_wind 366 non-null values
mean_wind 366 non-null values
min_wind 365 non-null values
precipitation 366 non-null values
cloud_cover 366 non-null values
events 366 non-null values
wind_dir 366 non-null values
rain 366 non-null values
thunderstorm 366 non-null values
fog 366 non-null values
snow 366 non-null values
dtypes: bool(4), float64(5), int64(16), object(1)
Our new columns show up at the bottom. We can access them now with the dot syntax.
data.rain
Output:
date
2012-03-10 False
2012-03-11 True
2012-03-12 True
2012-03-13 False
2012-03-14 False
2012-03-15 True
2012-03-16 False
2012-03-17 False
2012-03-18 True
2012-03-19 False
2012-03-20 False
2012-03-21 False
2012-03-22 True
2012-03-23 True
2012-03-24 True
...
2013-02-24 False
2013-02-25 False
2013-02-26 True
2013-02-27 False
2013-02-28 True
2013-03-01 False
2013-03-02 False
2013-03-03 False
2013-03-04 True
2013-03-05 True
2013-03-06 False
2013-03-07 False
2013-03-08 False
2013-03-09 False
2013-03-10 False
Freq: D, Name: rain, Length: 366
We can also do cool things like find out how many True values there are (i.e., how many days had rain)…
data.rain.sum()
Output:
121
…and get all the days that had both rain and snow!
data[data.rain & data.snow]
Output:
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 7 entries, 2012-11-12 00:00:00 to 2013-03-05 00:00:00
Data columns:
max_temp 7 non-null values
mean_temp 7 non-null values
min_temp 7 non-null values
max_dew 7 non-null values
mean_dew 7 non-null values
min_dew 7 non-null values
max_humidity 7 non-null values
mean_humidity 7 non-null values
min_humidity 7 non-null values
max_pressure 7 non-null values
mean_pressure 7 non-null values
min_pressure 7 non-null values
max_visibilty 7 non-null values
mean_visibility 7 non-null values
min_visibility 7 non-null values
max_wind 7 non-null values
mean_wind 7 non-null values
min_wind 7 non-null values
precipitation 7 non-null values
cloud_cover 7 non-null values
events 7 non-null values
wind_dir 7 non-null values
rain 7 non-null values
thunderstorm 7 non-null values
fog 7 non-null values
snow 7 non-null values
dtypes: bool(4), float64(5), int64(16), object(1)
Exercise 6:
Was the mean temperature more variable on days with rain and snow than on days with just rain or just snow?
Hint: don't forget the std() function
days_with_rain = data[data.rain == True]
days_with_snow = data[data.snow == True]
rain_std = days_with_rain.mean_temp.std()
snow_std = days_with_snow.mean_temp.std()
if rain_std > snow_std:
print "Rainy days were more variable"
elif snow_std > rain_std:
print "Snowy days were more variable"
else:
print "They were the same"
Output:
Rainy days were more variable
Plotting
We've already seen how the hist() function makes generating histograms a snap. Let's look at the plot() function now.
data.max_temp.plot()
Output:
<matplotlib.axes.AxesSubplot at 0x46f2610>
That one line of code did a lot for us. First, it created a nice looking line plot using the maximum temperature column from our DataFrame. Second, because we used datetime objects in our index, pandas labeled the x-axis appropriately.
Pandas is smart too. If we're only looking at a couple of days, the x-axis looks different:
data.max_temp.tail().plot()
Output:
<matplotlib.axes.AxesSubplot at 0x4c03710>
Prefer a bar plot? Pandas has got your covered.
data.max_temp.tail().plot(kind="bar", rot=10)
Output:
<matplotlib.axes.AxesSubplot at 0x4e803d0>
The plot() function returns a matplotlib AxesSubPlot object. You can pass this object into subsequent calls to plot() in order to compose plots.
Although plot() takes a variety of parameters to customize your plot, users familiar with matplotlib will feel right at home with the AxesSubPlot object.
ax = data.max_temp.plot(title="Min and Max Temperatures")
data.min_temp.plot(style="red", ax=ax)
ax.set_ylabel("Temperature (F)")
Output:
<matplotlib.text.Text at 0x50f0f90>
Exercise 7:
Add the mean temperature to the previous plot using a green line. Also, add a legend with the legend() method of ax.
ax = data.max_temp.plot(title="Min and Max Temperatures")
data.min_temp.plot(style="red", ax=ax)
data.mean_temp.plot(style="green", ax=ax)
ax.set_ylabel("Temperature (F)")
Output:
<matplotlib.text.Text at 0x532ec90>
Getting Data Out
Writing data out in pandas is as easy as getting data in. To save our DataFrame out to a new csv file, we can just do this:
data.to_csv("data/weather-mod.csv")
Want to make that tab separated instead? No problem.
data.to_csv("data/weather-mod.tsv", sep="\t")
There's also support for reading and writing Excel files, if you need it.
Miscellanea
We've only covered a small fraction of the pandas library here. Before I wrap up, however, there are a few miscellaneous tips I'd like to go over.
First, it can be confusing to know when an operation will modify a DataFrame and when it will return a copy to you. Pandas behavior here is entirely dictated by NumPy, and some situations are unintuitive.
For example, what do you think will happen here?
for idx, row in data.iterrows():
row["max_temp"] = 0
data.max_temp.head()
Output:
date
2012-03-10 56
2012-03-11 67
2012-03-12 71
2012-03-13 76
2012-03-14 80
Freq: D, Name: max_temp
Contrary to what you might expect, modifying row did not modify data!
This is because row is a copy, and does not point back to the original DataFrame.
Here's the right way to do it:
for idx, row in data.iterrows():
data.ix[idx, "max_temp"] = 0
any(data.max_temp != 0) # Any rows with max_temp not equal to zero?
Output:
False
Just to make you even more confused, this also doesn't work:
for idx, row in data.iterrows():
data.ix[idx]["max_temp"] = 100
data.max_temp.head()
Output:
date
2012-03-10 0
2012-03-11 0
2012-03-12 0
2012-03-13 0
2012-03-14 0
Freq: D, Name: max_temp
When using apply(), the default behavior is to go over columns.
data.apply(lambda c: c.name)
Output:
max_temp max_temp
mean_temp mean_temp
min_temp min_temp
max_dew max_dew
mean_dew mean_dew
min_dew min_dew
max_humidity max_humidity
mean_humidity mean_humidity
min_humidity min_humidity
max_pressure max_pressure
mean_pressure mean_pressure
min_pressure min_pressure
max_visibilty max_visibilty
mean_visibility mean_visibility
min_visibility min_visibility
max_wind max_wind
mean_wind mean_wind
min_wind min_wind
precipitation precipitation
cloud_cover cloud_cover
events events
wind_dir wind_dir
rain rain
thunderstorm thunderstorm
fog fog
snow snow
You can make apply() go over rows by passing axis=1
data.apply(lambda r: r["max_pressure"] - r["min_pressure"], axis=1)
Output:
date
2012-03-10 0.19
2012-03-11 0.24
2012-03-12 0.25
2012-03-13 0.15
2012-03-14 0.11
2012-03-15 0.11
2012-03-16 0.07
2012-03-17 0.11
2012-03-18 0.12
2012-03-19 0.11
2012-03-20 0.10
2012-03-21 0.12
2012-03-22 0.10
2012-03-23 0.27
2012-03-24 0.09
...
2013-02-24 0.15
2013-02-25 0.33
2013-02-26 0.40
2013-02-27 0.23
2013-02-28 0.28
2013-03-01 0.11
2013-03-02 0.11
2013-03-03 0.08
2013-03-04 0.20
2013-03-05 0.26
2013-03-06 0.53
2013-03-07 0.13
2013-03-08 0.13
2013-03-09 0.36
2013-03-10 0.05
Length: 366
When you call drop(), though, it's flipped. To drop a column, you need to pass axis=1
data.drop(["events"], axis=1).columns
Output:
Index([max_temp, mean_temp, min_temp, max_dew, mean_dew, min_dew,
max_humidity, mean_humidity, min_humidity, max_pressure,
mean_pressure, min_pressure, max_visibilty, mean_visibility,
min_visibility, max_wind, mean_wind, min_wind, precipitation,
cloud_cover, wind_dir, rain, thunderstorm, fog, snow], dtype=object)