Cheat Sheet: The pandas DataFrame Object I

Preliminaries Get your data into a DataFrame

Always start by importing these Python modules Instantiate an empty DataFrame

import numpy as np df = DataFrame()
import [Link] as plt
import pandas as pd Load a DataFrame from a CSV file
from pandas import DataFrame, Series df = pd.read_csv('[Link]')# often works
Note: these are the recommended import aliases df = pd.read_csv('[Link]', header=0,
Note: you can put these into a PYTHONSTARTUP file index_col=0, quotechar='"',sep=':',
na_values = ['na', '-', '.', ''])
Note: refer to pandas docs for all arguments
Cheat sheet conventions Get data from inline CSV text to a DataFrame
from io import StringIO
Code examples data = """, Animal, Cuteness, Desirable
# Code examples are found in yellow boxes row-1, dog, 8.7, True
# These are designed to be cut and paste row-2, cat, 9.5, True
row-3, bat, 2.6, False"""
df = pd.read_csv(StringIO(data),
In the code examples, typically I use: header=0, index_col=0,
• s to represent a pandas Series object; skipinitialspace=True)
• df to represent a pandas DataFrame object; Note: skipinitialspace=True allows a pretty layout
• idx to represent a pandas Index object.
• Also: t – tuple, l – list, b – Boolean, i – integer, Load DataFrames from a Microsoft Excel file
a – numpy array, st – string, d – dictionary, etc. # Each Excel sheet in a Python dictionary
workbook = [Link]('[Link]')
d = {} # start with an empty dictionary
for sheet_name in workbook.sheet_names:
The conceptual model df = [Link](sheet_name)
d[sheet_name] = df
DataFrame object: is a two-dimensional table of data Note: the parse() method takes many arguments like
with column and row indexes (something like a spread read_csv() above. Refer to the pandas documentation.
sheet). The columns are made up of Series objects.
Load a DataFrame from a MySQL database
Column index ([Link]) import pymysql
from sqlalchemy import create_engine
engine = create_engine('mysql+pymysql://'
Series of data

Series of data

Series of data

Series of data

Series of data

Series of data

Series of data

+'USER:PASSWORD@HOST/DATABASE')
([Link])

df = pd.read_sql_table('table', engine)
Row index

Data in Series then combine into a DataFrame

# Example 1 ...
s1 = Series(range(6))
s2 = s1 * s1
[Link] = [Link] + 2# misalign indexes
df = [Link]([s1, s2], axis=1)
A DataFrame has two Indexes:
• Typically, the column index ([Link]) is a list of # Example 2 ...
strings (observed variable names) or (less s3 = Series({'Tom':1, 'Dick':4, 'Har':9})
commonly) integers s4 = Series({'Tom':3, 'Dick':2, 'Mar':5})
• Typically, the row index ([Link]) might be: df = [Link]({'A':s3, 'B':s4 }, axis=1)
o Integers - for case or row numbers; Note: 1st method has in integer column labels
o Strings – for case names; or Note: 2nd method does not guarantee col order
o DatetimeIndex or PeriodIndex – for time series Note: index alignment on DataFrame creation

Series object: an ordered, one-dimensional array of Get a DataFrame from a Python dictionary
data with an index. All the data in a Series is of the # default --- assume data is in columns
same data type. Series arithmetic is vectorised after first df = DataFrame({
aligning the Series index for each of the operands. 'col0' : [1.0, 2.0, 3.0, 4.0],
'col1' : [100, 200, 300, 400]
s1 = Series(range(0,4)) # -> 0, 1, 2, 3 })
s2 = Series(range(1,5)) # -> 1, 2, 3, 4
s3 = s1 + s2 # -> 1, 3, 5, 7
Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
1
Get a DataFrame from data in a Python dictionary
# --- use helper method for data in rows Working with the whole DataFrame
df = DataFrame.from_dict({ # data by row
# rows as python dictionaries Peek at the DataFrame contents/structure
'row0' : {'col0':0, 'col1':'A'},
[Link]() # index & data types
'row1' : {'col0':1, 'col1':'B'}
dfh = [Link](i) # get first i rows
}, orient='index')
dft = [Link](i) # get last i rows
dfs = [Link]() # summary stats cols
df = DataFrame.from_dict({ # data by row
top_left_corner_df = [Link][:4, :4]
# rows as python lists
'row0' : [1, 1+1j, 'A'],
'row1' : [2, 2+2j, 'B'] DataFrame non-indexing attributes
}, orient='index') dfT = df.T # transpose rows and cols
l = [Link] # list row and col indexes
Create play/fake data (useful for testing) (r_idx, c_idex) = [Link] # from above
s = [Link] # Series column data types
# --- simple - default integer indexes
b = [Link] # True for empty DataFrame
df = DataFrame([Link](50,5))
i = [Link] # number of axes (it is 2)
t = [Link] # (row-count, column-count)
# --- with a time-stamp row index:
i = [Link] # row-count * column-count
df = DataFrame([Link](500,5))
a = [Link] # get a numpy array for df
[Link] = pd.date_range('1/1/2005',
periods=len(df), freq='M')
DataFrame utility methods
# --- with alphabetic row and col indexes df = [Link]() # copy a DataFrame
# and a "groupable" variable df = [Link]() # rank each col (default)
import string df = df.sort_values(by=col)
import random df = df.sort_values(by=[col1, col2])
r = 52 # note: min r is 1; max r is 52 df = df.sort_index()
c = 5 df = [Link](dtype) # type conversion
df = DataFrame([Link](r, c),
columns = ['col'+str(i) for i in DataFrame iteration methods
range(c)], [Link]()# (col-index, Series) pairs
index = list((string. ascii_uppercase+ [Link]() # (row-index, Series) pairs
string.ascii_lowercase)[0:r]))
df['group'] = list( # example ... iterating over columns
''.join([Link]('abcde') for (name, series) in [Link]():
for _ in range(r)) ) print('Col name: ' + str(name))
print('First value: ' +
str([Link][0]) + '\n')

Saving a DataFrame Maths on the whole DataFrame (not a complete list)

df = [Link]() # absolute values
Saving a DataFrame to a CSV file df = [Link](o) # add df, Series or value
df.to_csv('[Link]', encoding='utf-8') s = [Link]() # non NA/null values
df = [Link]() # (cols default axis)
df = [Link]() # (cols default axis)
Saving DataFrames to an Excel Workbook
df = [Link]() # (cols default axis)
from pandas import ExcelWriter df = [Link]() # 1st diff (col def axis)
writer = ExcelWriter('[Link]') df = [Link](o) # div by df, Series, value
df1.to_excel(writer,'Sheet1') df = [Link](o) # matrix dot product
df2.to_excel(writer,'Sheet2') s = [Link]() # max of axis (col def)
[Link]() s = [Link]() # mean (col default axis)
s = [Link]()# median (col default)
Saving a DataFrame to MySQL s = [Link]() # min of axis (col def)
import pymysql df = [Link](o) # mul by df Series val
from sqlalchemy import create_engine s = [Link]() # sum axis (cols default)
e = create_engine('mysql+pymysql://' + df = [Link](df > 0.5, other=[Link])
'USER:PASSWORD@HOST/DATABASE') Note: The methods that return a series default to
df.to_sql('TABLE',e, if_exists='replace') working on columns.
Note: if_exists à 'fail', 'replace', 'append'
DataFrame select/filter rows/cols on label values
Saving to Python objects df = [Link](items=['a', 'b']) # by col
d = df.to_dict() # to dictionary df = [Link](items=[5], axis=0)# by row
str = df.to_string() # to string df = [Link](like='x') # keep x in col
m = df.as_matrix() # to numpy matrix df = [Link](regex='x') # regex in col
df = [Link](lambda x: not x%5)#5th rows
Note: select takes a Boolean function, for cols: axis=1
Note: filter defaults to cols; select defaults to rows
Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
2
 Set column values set based on criteria
Working with Columns df['b']=df['a'].where(df['a']>0,other=0)
df['d']=df['a'].where(df.b!=0,other=df.c)
Get column index and labels Note: where other can be a Series or a scalar
idx = [Link] # get col index
label = [Link][0] # first col label Data type conversions
l = [Link]()# list col labels st = df['col'].astype(str)# Series dtype
a = [Link] # array col labels a = df['col'].values # numpy array
l = df['col'].tolist() # python list
Change column labels Note: useful dtypes for Series conversion: int, float, str
[Link](columns={'old1':'new1', Trap: index lost in conversion from Series to array or list
'old2':'new2'}, inplace=True)
Note: can rename multiple columns at once. Common column-wide methods/attributes
value = df['col'].dtype # type of data
Selecting columns value = df['col'].size # col dimensions
s = df['colName'] # select col to Series value = df['col'].count()# non-NA count
df = df[['colName']] # select col to df value = df['col'].sum()
df = df[['a','b']] # select 2 or more value = df['col'].prod()
df = df[['c','a','b']]# change col order value = df['col'].min()
s = df[[Link][0]] # select by number value = df['col'].max()
df = df[[Link][[0, 3, 4]] # by number value = df['col'].mean() # also median()
df = [[Link][:-1]] # all but last col value = df['col'].cov(df['col2'])
s = [Link]('c') # get col & drop from df s = df['col'].describe()
s = df['col'].value_counts()
Selecting columns with Python attributes
Find index label for min/max values in column
s = df.a # same as s = df['a']
# cannot create new columns by attribute label = df['col1'].idxmin()
df.existing_column = df.a / df.b label = df['col1'].idxmax()
df['new_column'] = df.a / df.b
Trap: column names must be valid identifiers. Common column element-wise methods
s = df['col'].isnull()
Adding new columns to a DataFrame s = df['col'].notnull() # not isnull()
s = df['col'].astype(float)
df['new_col'] = range(len(df))
s = df['col'].abs()
df['new_col'] = [Link]([Link],len(df))
s = df['col'].round(decimals=0)
df['random'] = [Link](len(df))
s = df['col'].diff(periods=1)
df['index_as_col'] = [Link]
s = df['col'].shift(periods=1)
df1[['b','c']] = df2[['e','f']]
s = df['col'].to_datetime()
df3 = [Link](other=df2)
s = df['col'].fillna(0) # replace NaN w 0
Trap: When adding a new column, only items from the s = df['col'].cumsum()
new series that have a corresponding index in the s = df['col'].cumprod()
DataFrame will be added. The receiving DataFrame is s = df['col'].pct_change(periods=4)
not extended to accommodate the new series. s = df['col'].rolling_sum(periods=4,
Trap: when adding a python list or numpy array, the window=4)
column will be added by integer position. Note: also rolling_min(), rolling_max(), and many more.
Swap column contents Append a column of row sums to a DataFrame
df[['B', 'A']] = df[['A', 'B']] df['Total'] = [Link](axis=1)
Note: also means, mins, maxs, etc.
Dropping (deleting) columns (mostly by label)
df = [Link]('col1', axis=1) Multiply every column in DataFrame by Series
[Link]('col1', axis=1, inplace=True) df = [Link](s, axis=0) # on matched rows
df = [Link](['col1','col2'], axis=1)
Note: also add, sub, div, etc.
s = [Link]('col') # drops from frame
del df['col'] # even classic python works
df = [Link]([Link][0], axis=1)#first Selecting columns with .loc, .iloc and .ix
df = [Link]([Link][-1:],axis=1)#last df = [Link][:, 'col1':'col2'] # inclusive
df = [Link][:, 0:2] # exclusive
Vectorised arithmetic on columns
df['proportion']=df['count']/df['total'] Get the integer position of a column index label
df['percent'] = df['proportion'] * 100.0 j = [Link].get_loc('col_name')

Apply numpy mathematical functions to columns Test if column index values are unique/monotonic
df['log_data'] = [Link](df['col1']) if [Link].is_unique: pass # ...
Note: many many more numpy math functions b = [Link].is_monotonic_increasing
Hint: Prefer pandas math over numpy where you can. b = [Link].is_monotonic_decreasing

Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
3
 Select a slice of rows by label/index
Working with rows [inclusive-from : inclusive–to [ : step]]
df = df['a':'c'] # rows 'a' through 'c'
Get the row index and labels Trap: doesn't work on integer labelled rows
idx = [Link] # get row index
label = [Link][0] # first row label Append a row of column totals to a DataFrame
label = [Link][-1] # last row label # Option 1: use dictionary comprehension
l = [Link]() # get as a list sums = {col: df[col].sum() for col in df}
a = [Link] # get as an array sums_df = DataFrame(sums,index=['Total'])
df = [Link](sums_df)
Change the (row) index
[Link] = idx # new ad hoc index # Option 2: All done with pandas
df = df.set_index('A')# col A new index df = [Link](DataFrame([Link](),
df = df.set_index(['A', 'B'])# MultiIndex columns=['Total']).T)
df = df.reset_index() # replace old w new
# note: old index stored as a col in df Iterating over DataFrame rows
[Link] = range(len(df)) # set with list for (index, row) in [Link](): # pass
df = [Link](index=range(len(df))) Trap: row data type may be coerced.
df = df.set_index(keys=['r1','r2','etc'])
[Link](index={'old':'new'}, Sorting DataFrame rows values
inplace=True) df = [Link]([Link][0],
ascending=False)
Adding rows [Link](['col1', 'col2'], inplace=True)
df = original_df.append(more_rows_in_df)
Hint: convert row to a DataFrame and then append. Sort DataFrame by its row index
Both DataFrames should have same column labels. df.sort_index(inplace=True) # sort by row
df = df.sort_index(ascending=False)
Dropping rows (by name)
df = [Link]('row_label') Random selection of rows
df = [Link](['row1','row2']) # multi-row import random as r
k = 20 # pick a number
Boolean row selection by values in a column selection = [Link](range(len(df)), k)
df = df[df['col2'] >= 0.0] df_sample = [Link][selection, :]
df = df[(df['col3']>=1.0) | Note: this sample is not sorted
(df['col1']<0.0)]
df = df[df['col'].isin([1,2,5,7,11])] Drop duplicates in the row index
df = df[~df['col'].isin([1,2,5,7,11])] df['index'] = [Link] # 1 create new col
df = df[df['col'].[Link]('hello')] df = df.drop_duplicates(cols='index',
Trap: bitwise "or", "and" “not; (ie. | & ~) co-opted to be take_last=True)# 2 use new col
Boolean operators on a Series of Boolean del df['index'] # 3 del the col
Trap: need parentheses around comparisons. df.sort_index(inplace=True)# 4 tidy up

Selecting rows using isin over multiple columns Test if two DataFrames have same row index
# fake up some data len(a)==len(b) and all([Link]==[Link])
data = {1:[1,2,3], 2:[1,4,9], 3:[1,8,27]}
df = DataFrame(data) Get the integer position of a row or col index label
i = [Link].get_loc('row_label')
# multi-column isin
lf = {1:[1, 3], 3:[8, 27]} # look for Trap: index.get_loc() returns an integer for a unique
f = df[df[list(lf)].isin(lf).all(axis=1)] match. If not a unique match, may return a slice/mask.

Selecting rows using an index Get integer position of rows that meet condition
idx = df[df['col'] >= 2].index a = [Link](df['col'] >= 2) #numpy array
print([Link][idx])
Test if the row index values are unique/monotonic
Select a slice of rows by integer position if [Link].is_unique: pass # ...
[inclusive-from : exclusive-to [: step]] b = [Link].is_monotonic_increasing
default start is 0; default end is len(df) b = [Link].is_monotonic_decreasing
df = df[:] # copy DataFrame
df = df[0:2] # rows 0 and 1 Find row index duplicates
df = df[-1:] # the last row if [Link].has_duplicates:
df = df[2:3] # row 2 (the third row) print([Link]())
df = df[:-1] # all but the last row Note: also similar for column label duplicates.
df = df[::2] # every 2nd row (0 2 ..)
Trap: a single integer without a colon is a column label
for integer numbered columns.
Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
4
 Working with cells Summary: selecting using the DataFrame index

Selecting a cell by row and column labels Using the DataFrame index to select columns
value = [Link]['row', 'col'] s = df['col_label'] # returns Series
value = [Link]['row', 'col'] df = df[['col_label']]# return DataFrame
value = df['col'].at['row'] # tricky df = df[['L1', 'L2']] # select with list
Note: .at[] fastest label based scalar lookup df = df[index] # select with index
df = df[s] #select with Series
Setting a cell by row and column labels Note: the difference in return type with the first two
[Link]['row', 'col'] = value examples above based on argument type (scalar vs list).
[Link]['row', 'col'] = value
df['col'].at['row'] = value # tricky Using the DataFrame index to select rows
df = df['from':'inc_to']# label slice
Selecting and slicing on labels df = df[3:7] # integer slice
df = [Link]['row1':'row3', 'col1':'col3'] df = df[df['col'] > 0.5]# Boolean Series
df = [Link]['label'] # single label
Note: the "to" on this slice is inclusive.
df = [Link][container] # lab list/Series
df = [Link]['from':'to']# inclusive slice
Setting a cross-section by labels df = [Link][bs] # Boolean Series
[Link]['A':'C', 'col1':'col3'] = [Link] df = [Link][0] # single integer
[Link][1:2,'col1':'col2']=[Link]((2,2)) df = [Link][container] # int list/Series
[Link][1:2,'A':'C']=[Link][1:2,'A':'C'] df = [Link][0:5] # exclusive slice
Remember: inclusive "to" in the slice df = [Link][x] # loc then iloc

Selecting a cell by integer position Using the DataFrame index to select a cross-section
value = [Link][9, 3] # [row, col] # r and c can be scalar, list, slice
value = [Link][0, 0] # [row, col] [Link][r, c] # label accessor (row, col)
value = [Link][len(df)-1, [Link][r, c]# integer accessor
len([Link])-1] [Link][r, c] # label access int fallback
df[c].iloc[r]# chained – also for .loc
Selecting a range of cells by int position
df = [Link][2:4, 2:4] # subset of the df Using the DataFrame index to select a cell
df = [Link][:5, :5] # top left corner # r and c must be label or integer
s = [Link][5, :] # returns row as Series [Link][r, c] # fast scalar label accessor
df = [Link][5:6, :] # returns row as row [Link][r, c] # fast scalar int accessor
Note: exclusive "to" – same as python list slicing. df[c].iat[r] # chained – also for .at

Setting cell by integer position DataFrame indexing methods

[Link][0, 0] = value # [row, col] v = df.get_value(r, c) # get by row, col
[Link][7, 8] = value df = df.set_value(r,c,v)# set by row, col
df = [Link](key, axis) # get cross-section
Setting cell range by integer position df = [Link](items, like, regex, axis)
[Link][0:3, 0:5] = value df = [Link](crit, axis)
[Link][1:3, 1:4] = [Link]((2, 3))
[Link][1:3, 1:4] = [Link]((2, 3)) Note: the indexing attributes (.loc, .iloc, .ix, .at .iat) can
[Link][1:3, 1:4] = [Link]([[1, 1, 1], be used to get and set values in the DataFrame.
[2, 2, 2]]) Note: the .loc, iloc and .ix indexing attributes can accept
Remember: exclusive-to in the slice python slice objects. But .at and .iat do not.
Note: .loc can also accept Boolean Series arguments
.ix for mixed label and integer position indexing Avoid: chaining in the form df[col_indexer][row_indexer]
value = [Link][5, 'col1'] Trap: label slices are inclusive, integer slices exclusive.
df = [Link][1:5, 'col1':'col3']
Some index attributes and methods
Views and copies # --- some Index attributes
From the manual: Setting a copy can cause subtle b = idx.is_monotonic_decreasing
errors. The rules about when a view on the data is b = idx.is_monotonic_increasing
returned are dependent on NumPy. Whenever an array b = idx.has_duplicates
of labels or a Boolean vector are involved in the indexing i = [Link] # num of index levels
operation, the result will be a copy. # --- some Index methods
idx = [Link](dtype)# change data type
b = [Link](o) # check for equality
idx = [Link](o) # union of two indexes
i = [Link]() # number unique labels
label = [Link]() # minimum label
label = [Link]() # maximum label

Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
5
 Grouping
Joining/Combining DataFrames gb = [Link]('cat') # by one columns
gb = [Link](['c1','c2']) # by 2 cols
Three ways to join two DataFrames: gb = [Link](level=0) # multi-index gb
gb = [Link](level=['a','b']) # mi gb
• merge (a database/SQL-like join operation)
print([Link])
• concat (stack side by side or one on top of the other)
Note: groupby() returns a pandas groupby object
• combine_first (splice the two together, choosing
Note: the groupby object attribute .groups contains a
values from one over the other)
dictionary mapping of the groups.
Trap: NaN values in the group key are automatically
Merge on indexes
dropped – there will never be a NA group.
df_new = [Link](left=df1, right=df2,
how='outer', left_index=True,
Iterating groups – usually not needed
right_index=True)
for name, group in gb:
How: 'left', 'right', 'outer', 'inner'
print (name)
How: outer=union/all; inner=intersection print (group)
Merge on columns
Selecting a group
df_new = [Link](left=df1, right=df2,
dfa = [Link]('cat').get_group('a')
how='left', left_on='col1',
dfb = [Link]('cat').get_group('b')
right_on='col2')
Trap: When joining on columns, the indexes on the
passed DataFrames are ignored. Applying an aggregating function
Trap: many-to-many merges on a column can result in # apply to a column ...
an explosion of associated data. s = [Link]('cat')['col1'].sum()
s = [Link]('cat')['col1'].agg([Link])
# apply to the every column in DataFrame
Join on indexes (another way of merging)
s = [Link]('cat').agg([Link])
df_new = [Link](other=df2, on='col1', df_summary = [Link]('cat').describe()
how='outer') df_row_1s = [Link]('cat').head(1)
df_new = [Link](other=df2,on=['a','b'],
how='outer') Note: aggregating functions reduce the dimension by
one – they include: mean, sum, size, count, std, var,
Note: [Link]() joins on indexes by default.
sem, describe, first, last, min, max
[Link]() joins on common columns by
default.
Applying multiple aggregating functions
gb = [Link]('cat')
Simple concatenation is often the best
df=[Link]([df1,df2],axis=0)#top/bottom # apply multiple functions to one column
df = [Link]([df2, df3]) #top/bottom dfx = gb['col2'].agg([[Link], [Link]])
df=[Link]([df1,df2],axis=1)#left/right # apply to multiple fns to multiple cols
Trap: can end up with duplicate rows or cols dfy = [Link]({
Note: concat has an ignore_index parameter 'cat': np.count_nonzero,
'col1': [[Link], [Link], [Link]],
Combine_first 'col2': [[Link], [Link]]
df = df1.combine_first(other=df2) })
Note: gb['col2'] above is shorthand for
# multi-combine with python reduce() [Link]('cat')['col2'], without the need for regrouping.
df = reduce(lambda x, y:
x.combine_first(y), Transforming functions
[df1, df2, df3, df4, df5]) # transform to group z-scores, which have
Uses the non-null values from df1. The index of the # a group mean of 0, and a std dev of 1.
combined DataFrame will be the union of the indexes zscore = lambda x: ([Link]())/[Link]()
from df1 and df2. dfz = [Link]('cat').transform(zscore)

# replace missing data with group mean

mean_r = lambda x: [Link]([Link]())
Groupby: Split-Apply-Combine dfm = [Link]('cat').transform(mean_r)
Note: can apply multiple transforming functions in a
The pandas "groupby" mechanism allows us to split the manner similar to multiple aggregating functions above,
data into groups, apply a function to each group
independently and then combine the results. Applying filtering functions
Filtering functions allow you to make selections based
on whether each group meets specified criteria
# select groups with more than 10 members
eleven = lambda x: (len(x['col1']) >= 11)
df11 = [Link]('cat').filter(eleven)

Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
6
Group by a row index (non-hierarchical index)
df = df.set_index(keys='cat') Working with dates, times and their indexes
s = [Link](level=0)['col1'].sum()
dfg = [Link](level=0).sum() Dates and time – points and spans
With its focus on time-series data, pandas has a suite of
tools for managing dates and time: either as a point in
time (a Timestamp) or as a span of time (a Period).
Pivot Tables: working with long and wide data
t = [Link]('2013-01-01')
t = [Link]('2013-01-01 [Link]')
These features work with and often create t = [Link]('2013-01-01 [Link].7')
hierarchical or multi-level Indexes; p = [Link]('2013-01-01', freq='M')
(the pandas MultiIndex is powerful and complex). Note: Timestamps should be in range 1678 and 2261
years. (Check [Link] and [Link]).
Pivot, unstack, stack and melt
Pivot tables move from long format to wide format data A Series of Timestamps or Periods
# Let's start with data in long format ts = ['2015-04-01 [Link]',
from StringIO import StringIO # python2.7 '2014-04-02 [Link]']
#from io import StringIO # python 3
data = """Date,Pollster,State,Party,Est # Series of Timestamps (good)
13/03/2014, Newspoll, NSW, red, 25 s = pd.to_datetime([Link](ts))
13/03/2014, Newspoll, NSW, blue, 28
13/03/2014, Newspoll, Vic, red, 24 # Series of Periods (often not so good)
13/03/2014, Newspoll, Vic, blue, 23 s = [Link]( [[Link](x, freq='M')
13/03/2014, Galaxy, NSW, red, 23 for x in ts] )
13/03/2014, Galaxy, NSW, blue, 24 s = [Link](
13/03/2014, Galaxy, Vic, red, 26 [Link](ts,freq='S'))
13/03/2014, Galaxy, Vic, blue, 25 Note: While Periods make a very useful index; they may
13/03/2014, Galaxy, Qld, red, 21 be less useful in a Series.
13/03/2014, Galaxy, Qld, blue, 27"""
df = pd.read_csv(StringIO(data), From non-standard strings to Timestamps
header=0, skipinitialspace=True) t = ['[Link].7654-JAN092002',
'[Link].6589-FEB082016']
# pivot to wide format on 'Party' column s = [Link](pd.to_datetime(t,
# 1st: set up a MultiIndex for other cols format="%H:%M:%S.%f-%b%d%Y"))
df1 = df.set_index(['Date', 'Pollster', Also: %B = full month name; %m = numeric month;
'State']) %y = year without century; and more …
# 2nd: do the pivot
wide1 = [Link](columns='Party') Dates and time – stamps and spans as indexes
An index of Timestamps is a DatetimeIndex.
# unstack to wide format on State / Party
An index of Periods is a PeriodIndex.
# 1st: MultiIndex all but the Values col
df2 = df.set_index(['Date', 'Pollster', date_strs = ['2014-01-01', '2014-04-01',
'State', 'Party']) '2014-07-01', '2014-10-01']
# 2nd: unstack a column to go wide on it
wide2 = [Link]('State') dti = [Link](date_strs)
wide3 = [Link]() # pop last index
pid = [Link](date_strs, freq='D')
# Use stack() to get back to long format pim = [Link](date_strs, freq='M')
long1 = [Link]() piq = [Link](date_strs, freq='Q')
# Then use reset_index() to remove the
# MultiIndex. print (pid[1] - pid[0]) # 90 days
long2 = long1.reset_index() print (pim[1] - pim[0]) # 3 months
print (piq[1] - piq[0]) # 1 quarter
# Or melt() back to long format
# 1st: flatten the column index time_strs = ['2015-01-01 [Link].12345',
[Link] = ['_'.join(col).strip() '2015-01-01 [Link].67890']
for col in [Link]] pis = [Link](time_strs, freq='U')
# 2nd: remove the MultiIndex
wdf = wide1.reset_index() [Link] = pd.period_range('2015-01',
# 3rd: melt away periods=len(df), freq='M')
long3 = [Link](wdf, value_vars=
['Est_blue', 'Est_red'], dti = pd.to_datetime(['04-01-2012'],
var_name='Party', id_vars=['Date', dayfirst=True) # Australian date format
'Pollster', 'State']) pi = pd.period_range('1960-01-01',
'2015-12-31', freq='M')
Note: See documentation, there are many arguments to
these methods. Hint: unless you are working in less than seconds,
prefer PeriodIndex over DateTimeImdex.

Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
7
Period frequency constants (not a complete list) Upsampling and downsampling
Name Description # upsample from quarterly to monthly
U Microsecond pi = pd.period_range('1960Q1',
L Millisecond periods=220, freq='Q')
S Second df = DataFrame([Link](len(pi),5),
index=pi)
T Minute
dfm = [Link]('M', convention='end')
H Hour # use ffill or bfill to fill with values
D Calendar day
B Business day # downsample from monthly to quarterly
W-{MON, TUE, …} Week ending on … dfq = [Link]('Q', how='sum')
MS Calendar start of month
M Calendar end of month Time zones
QS-{JAN, FEB, …} Quarter start with year starting t = ['2015-06-30 [Link]',
(QS – December) '2015-12-31 [Link]']
Q-{JAN, FEB, …} Quarter end with year ending (Q dti = pd.to_datetime(t
– December) ).tz_localize('Australia/Canberra')
AS-{JAN, FEB, …} Year start (AS - December) dti = dti.tz_convert('UTC')
ts = [Link]('now',
A-{JAN, FEB, …} Year end (A - December) tz='Europe/London')
From DatetimeIndex to Python datetime objects # get a list of all time zones
dti = [Link](pd.date_range( import pyzt
start='1/1/2011', periods=4, freq='M')) for tz in pytz.all_timezones:
s = Series([1,2,3,4], index=dti) print tz
na = dti.to_pydatetime() #numpy array Note: by default, Timestamps are created without time
na = [Link].to_pydatetime() #numpy array zone information.

Frome Timestamps to Python dates or times Row selection with a time-series index
df['date'] = [[Link]() for x in df['TS']] # start with the play data above
df['time'] = [[Link]() for x in df['TS']] idx = pd.period_range('2015-01',
Note: converts to [Link] or [Link]. But periods=len(df), freq='M')
does not convert to [Link]. [Link] = idx

From DatetimeIndex to PeriodIndex and back february_selector = ([Link] == 2)

df = DataFrame([Link](20,3)) february_data = df[february_selector]
[Link] = pd.date_range('2015-01-01',
periods=len(df), freq='M') q1_data = df[([Link] >= 1) &
dfp = df.to_period(freq='M') ([Link] <= 3)]
dft = dfp.to_timestamp()
Note: from period to timestamp defaults to the point in mayornov_data = df[([Link] == 5)
| ([Link] == 11)]
time at the start of the period.
totals = [Link]([Link]).sum()
Working with a PeriodIndex
Also: year, month, day [of month], hour, minute, second,
pi = pd.period_range('1960-01','2015-12',
dayofweek [Mon=0 .. Sun=6], weekofmonth, weekofyear
freq='M')
na = [Link] # numpy array of integers [numbered from 1], week starts on Monday], dayofyear
lp = [Link]() # python list of Periods [from 1], …
sp = Series(pi)# pandas Series of Periods
ss = Series(pi).astype(str) # S of strs The [Link] accessor attribute
ls = Series(pi).astype(str).tolist() DataFrame columns that contain datetime-like objects
can be manipulated with the .dt accessor attribute
Get a range of Timestamps t = ['2012-04-14 [Link].307000',
dr = pd.date_range('2013-01-01', '2011-05-14 [Link].457000',
'2013-12-31', freq='D') '2010-06-14 [Link].520000']

# a Series of time stamps

Error handling with dates s = [Link](pd.to_datetime(t))
# 1st example returns string not Timestamp print([Link]) # datetime64[ns]
t = pd.to_datetime('2014-02-30') print([Link]) # 56, 24, 7
# 2nd example returns NaT (not a time) print([Link]) # 4, 5, 6
t = pd.to_datetime('2014-02-30',
coerce=True) # a Series of time periods
# NaT like NaN tests True for isnull() s = [Link]([Link](t,freq='Q'))
b = [Link](t) # --> True print([Link]) # datetime64[ns]
print([Link]) # 2, 2, 2
The tail of a time-series DataFrame print([Link]) # 2012, 2011, 2010
df = [Link]("5M") # the last five months
Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
8
 Bar plots
Plotting from the DataFrame bins = [Link](-10,15,26)
binned = [Link]()
Import matplotlib, choose a matplotlib style for x in [Link]:
y=[Link](df[x],bins,labels=bins[:-1])
import [Link] as plt
y=y.value_counts().sort_index()
print([Link])
binned = [Link]([binned,y],axis=1)
[Link]('ggplot')
[Link] = [Link](float)
[Link] += ([Link](bins) / 2.0)
Fake up some data (which we reuse repeatedly) ax = [Link](stacked=False,
a = [Link](0,1,999) width=0.8) # for bar width
b = [Link](1,2,999)
c = [Link](2,3,999)
df = [Link]([a,b,c]).T
[Link] =['A', 'B', 'C']

Line plot
df1 = [Link]()
ax = [Link]()

# from here down – standard plot output

Horizontal bars
ax.set_title('Title')
ax.set_xlabel('X Axis') ax = binned['A'][([Link] >= -4) &
ax.set_ylabel('Y Axis') ([Link] <= 4)].[Link]()
fig = [Link]
fig.set_size_inches(8, 3)
fig.tight_layout(pad=1)
[Link]('[Link]', dpi=125)

[Link](fig)

Density plot
ax = [Link]()

Box plot
ax = [Link](vert=False)
# followed by the same code as above

Scatter plot
ax = [Link](x='A', y='C')

ax = [Link](column='c1', by='c2')

Histogram
ax = df['A'].[Link](bins=20) Pie chart
s = [Link](data=[10, 20, 30],
index = ['dogs', 'cats', 'birds'])
ax = [Link](autopct='%.1f')

ax.set_title('Pie Chart')
ax.set_aspect(1) # make it round
ax.set_ylabel('') # remove default

fig = [Link]
fig.set_size_inches(8, 3)
Multiple histograms
[Link]('[Link]', dpi=125)
ax = [Link](bins=25, alpha=0.5)
ax = [Link](bins=25, stacked=True) [Link](fig)
Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
9
 Change the range plotted
ax.set_xlim([-5, 5])
lower, upper = ax.get_ylim()
ax.set_ylim([lower-1, upper+1])

Add a footnote
# after the fig.tight_layout(pad=1)
[Link](0.99, 0.01, 'Footnote',
A line and bar on the same chart ha='right', va='bottom',
In matplotlib, bar charts visualise categorical or discrete fontsize='x-small',
data. Line charts visualise continuous data. This makes fontstyle='italic', color='#999999')
it hard to get bars and lines on the same chart. Typically
combined charts either have too many labels, and/or the
lines and bars are misaligned or missing. You need to Working with missing and non-finite data
trick matplotlib a bit … pandas makes this tricking easier
# start with fake percentage growth data Working with missing data
s = [Link]([Link]( Pandas uses the not-a-number construct ([Link] and
1.02, 0.015, 40)) float('nan')) to indicate missing data. The Python None
s = [Link]() can arise in data as well. It is also treated as missing
dfg = ([Link]([s / [Link](1),
data; as is the pandas not-a-time construct
s / [Link](4)], axis=1) * 100) - 100
([Link]).
[Link] = ['Quarter', 'Annual']
[Link] = pd.period_range('2010-Q1',
periods=len(dfg), freq='Q') Missing data in a Series
s = Series( [8,None,float('nan'),[Link]])
# reindex with integers from 0; keep old #[8, NaN, NaN, NaN]
old = [Link] [Link]() #[False, True, True, True]
[Link] = range(len(dfg)) [Link]()#[True, False, False, False]
[Link](0)#[8, 0, 0, 0]
# plot the line from pandas
ax = dfg['Annual'].plot(color='blue', Missing data in a DataFrame
label='Year/Year Growth') df = [Link]() # drop all rows with NaN
df = [Link](axis=1) # same for cols
# plot the bars from pandas df=[Link](how='all') #drop all NaN row
dfg['Quarter'].[Link](ax=ax, df=[Link](thresh=2) # drop 2+ NaN in r
label='Q/Q Growth', width=0.8) # only drop row if NaN in a specified col
df = [Link](df['col'].notnull())
# relabel the x-axis more appropriately
ticks = [Link][(([Link]+0)%4)==0] Recoding missing data
labs = [Link](old[ticks]).astype(str)
[Link](0, inplace=True) # [Link] à 0
ax.set_xticks(ticks)
s = df['col'].fillna(0) # [Link] à 0
ax.set_xticklabels([Link]('Q',
df = [Link](r'\s+', [Link],
'\nQ'), rotation=0)
regex=True) # white space à [Link]
# fix the range of the x-axis … skip 1st
ax.set_xlim([0.5,len(dfg)-0.5]) Non-finite numbers
With floating point numbers, pandas provides for
# add the legend positive and negative infinity.
l=[Link](loc='best',fontsize='small') s = Series([float('inf'), float('-inf'),
[Link], -[Link]])
# finish off and plot in the usual manner Pandas treats integer comparisons with plus or minus
ax.set_title('Fake Growth Data') infinity as expected.
ax.set_xlabel('Quarter')
ax.set_ylabel('Per cent') Testing for finite numbers
(using the data from the previous example)
fig = [Link]
b = [Link](s)
fig.set_size_inches(8, 3)
fig.tight_layout(pad=1)
[Link]('[Link]', dpi=125)

[Link](fig)

Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
10
 Working with Categorical Data Working with strings

Categorical data Working with strings

The pandas Series has an R factors-like data type for # assume that df['col'] is series of
encoding categorical data. strings
s = Series(['a','b','a','c','b','d','a'], s = df['col'].[Link]()
dtype='category') s = df['col'].[Link]()
df['B'] = df['A'].astype('category') s = df['col'].[Link]()
Note: the key here is to specify the "category" data type.
Note: categories will be ordered on creation if they are # the next set work like Python
sortable. This can be turned off. See ordering below. df['col'] += 'suffix' # append
df['col'] *= 2 # duplicate
Convert back to the original data type s = df['col1'] + df['col2'] # concatenate
s = Series(['a','b','a','c','b','d','a'], Most python string functions are replicated in the pandas
dtype='category') DataFrame and Series objects.
s = [Link]('string')
Regular expressions
Ordering, reordering and sorting s = df['col'].[Link]('regex')
s = df['col'].[Link]('regex')
s = Series(list('abc'), dtype='category')
s = df['col'].[Link]('regex')
print ([Link])
s = df['col'].[Link]('old', 'new')
s=[Link].reorder_categories(['b','c','a'])
df['b'] = [Link]('(pattern)')
s = [Link]()
[Link] = False Note: pandas has many more regex methods.
Trap: category must be ordered for it to be sorted

Renaming categories
s = Series(list('abc'), dtype='category')
[Link] = [1, 2, 3] # in place
s = [Link].rename_categories([4,5,6])
# using a comprehension ...
[Link] = ['Group ' + str(i)
for i in [Link]]
Trap: categories must be uniquely named

Adding new categories

s = [Link].add_categories([4])

Removing categories
s = [Link].remove_categories([4])
[Link].remove_unused_categories() #inplace

Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
11
 Basic Statistics

Summary statistics
s = df['col1'].describe()
df1 = [Link]()

DataFrame – key stats methods

[Link]() # pairwise correlation cols
[Link]() # pairwise covariance cols
[Link]() # kurtosis over cols (def)
[Link]() # mean absolute deviation
[Link]() # standard error of mean
[Link]() # variance over cols (def)

Value counts
s = df['col1'].value_counts()

Cross-tabulation (frequency count)

ct = [Link](index=df['a'],
cols=df['b'])

Quantiles and ranking

quants = [0.05, 0.25, 0.5, 0.75, 0.95]
q = [Link](quants)
r = [Link]()

Histogram binning
count, bins = [Link](df['col1'])
count, bins = [Link](df['col'],
bins=5)
count, bins = [Link](df['col1'],
bins=[-3,-2,-1,0,1,2,3,4])

Regression
import [Link] as sm
result = [Link](formula="col1 ~ col2 +
col3", data=df).fit()
print ([Link])
print ([Link]())

Smoothing example using rolling_apply

k3x5 = [Link]([1,2,3,3,3,2,1]) / 15.0
s = pd.rolling_apply(df['col1'],
window=7,
func=lambda x: (x * k3x5).sum(),
min_periods=7, center=True)

Cautionary note

This cheat sheet was cobbled together by bots roaming

the dark recesses of the Internet seeking ursine and
pythonic myths. There is no guarantee the narratives
were captured and transcribed accurately. You use
these notes at your own risk. You have been warned.

Version: This cheat sheet was last updated with

Python 3.5 and pandas 0.19.2 in mind.

Errors: If you find any errors, please email me at

markthegraph@[Link]; (but please do not correct
my use of Australian-English spelling conventions).

Version 15 January 2017 - [Draft – Mark Graph – mark dot the dot graph at gmail dot com – @Mark_Graph on twitter]
12