Taller de manejo de Datos en Pandas

¿Qué es?

Pandas es una librería de código abierto muy popular dentro del ámbito de Data Science y Machine Learning puesto que facilita el tratamiento y manipulación de datos.

Pandas nació de la necesidad de tener todas las funcionalidades necesarias para un analista de datos en una sola librería:

import pandas as pd

¿Cómo importar datos?

Para leer los datos la estructura básica es pd.read_tipo-archivo. Si estamos en colab y queremos utilizar algunos datos que están en nuestro google drive podemos utilizar:

      from google.colab import drive
      drive.mount('/content/gdrive')

Primeros pasos Pandas

#from google.colab import ##drive correr solo en colab
#drive.mount('/content/gdrive')
counties = pd.read_excel("https://github.com/AprendizajeProfundo/Libro_Fundamentos_Programacion/blob/main/Python/Datos/counties.xlsx?raw=true")#cambiar dirección

Para tener una idea de qué variables y datos tengo en lo base de datos, sólo debemos llamarla.

counties

	codestate	codecounty	county	population	area
0	1	1001	Auta#%&()uga	54571.0	594.436000
1	1	1003	Baldwin#%&() ?	182265.0	1589.784000
2	1	1005	Barbour	27457.0	884.876000
3	1	1007	Bi#%&()bb	22915.0	622.582000
4	1	1009	Blount ?	57322.0	644.776000
...	...	...	...	...	...
3229	72	72151	Yabucoa	37941.0	55.215000
3230	72	72153	Ya_uco	42043.0	68.192000
3231	78	78010	; St. Croix ?	50601.0	83.345868
3232	78	78020	St. John	4170.0	19.689867
3233	78	78030	St. Thomas	51634.0	31.313503

3234 rows × 5 columns

En lo anterior veiamos todas las variables con las 5 primeras y 5 últimas filas. Pero, podemos ver una cantidad determinada de registros tanto del inicio de la tabla como del final. Utlizando .head() y .tail() respectivamente.

counties.head(3) #por defecto salen 5 primeros

	codestate	codecounty	county	population	area
0	1	1001	Auta#%&()uga	54571.0	594.436
1	1	1003	Baldwin#%&() ?	182265.0	1589.784
2	1	1005	Barbour	27457.0	884.876

counties.tail(2)

	codestate	codecounty	county	population	area
3232	78	78020	St. John	4170.0	19.689867
3233	78	78030	St. Thomas	51634.0	31.313503

Si aparte de ver solamente la tabla, queremos más información de la base de datos, por ejemplo sus dimensiones y tipo de variables podemos utilizar lo siguiente:

counties.info()
print("\n")
print("La forma de la base de datos es:",counties.shape)  #shape:forma de un array

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3234 entries, 0 to 3233
Data columns (total 5 columns):
 #   Column      Non-Null Count  Dtype  
---  ------      --------------  -----  
 0   codestate   3234 non-null   int64  
 1   codecounty  3234 non-null   int64  
 2   county      3234 non-null   object 
 3   population  3232 non-null   float64
 4   area        3234 non-null   float64
dtypes: float64(2), int64(2), object(1)
memory usage: 126.5+ KB


La forma de la base de datos es: (3234, 5)

dtypes

print(counties.dtypes) #tipo de objeto en cada columna
print("\n")
print(counties.describe())
print("\n")
print(counties.describe(include="all")) #summary, si no se pone el include solo aparece de las variables numéricas

codestate       int64
codecounty      int64
county         object
population    float64
area          float64
dtype: object


         codestate    codecounty    population           area
count  3234.000000   3234.000000  3.232000e+03    3234.000000
mean     31.441868  31544.737786  9.679656e+04    1093.361817
std      16.411236  16425.545223  3.088044e+05    3564.706999
min       1.000000   1001.000000  1.700000e+01       0.031696
25%      19.000000  19039.500000  1.129700e+04     416.360000
50%      30.000000  30038.000000  2.607550e+04     602.977500
75%      46.000000  46128.500000  6.566050e+04     913.884500
max      78.000000  78030.000000  9.818605e+06  145504.789000


          codestate    codecounty    county    population           area
count   3234.000000   3234.000000      3234  3.232000e+03    3234.000000
unique          NaN           NaN      2625           NaN            NaN
top             NaN           NaN  Lincoln            NaN            NaN
freq            NaN           NaN        14           NaN            NaN
mean      31.441868  31544.737786       NaN  9.679656e+04    1093.361817
std       16.411236  16425.545223       NaN  3.088044e+05    3564.706999
min        1.000000   1001.000000       NaN  1.700000e+01       0.031696
25%       19.000000  19039.500000       NaN  1.129700e+04     416.360000
50%       30.000000  30038.000000       NaN  2.607550e+04     602.977500
75%       46.000000  46128.500000       NaN  6.566050e+04     913.884500
max       78.000000  78030.000000       NaN  9.818605e+06  145504.789000

Seleccionar subconjuntos de una base de datos

Documentación

Escoger columnas

Para ver el nombre de las columnas que existen, podemos utilizar .columns.

counties.columns

Index(['codestate', 'codecounty', 'county', 'population', 'area'], dtype='object')

Escoger una columna específica o un conjunto de ellas

condado = counties["county"]
condado

0               Auta#%&()uga 
1        Baldwin#%&()  ?     
2                    Barbour 
3                  Bi#%&()bb 
4              Blount  ?     
                ...          
3229                 Yabucoa 
3230                  Ya_uco 
3231       ; St. Croix ?     
3232                 St. John
3233               St. Thomas
Name: county, Length: 3234, dtype: object

print(condado.shape) #array unidimensional con 3234 elementos
print(counties["county"].shape)
print(type(condado))

(3234,)
(3234,)
<class 'pandas.core.series.Series'>

codi_condado=counties[["codecounty","county"]]
print(codi_condado)
print(type(codi_condado))

      codecounty                 county
0           1001          Auta#%&()uga 
1           1003   Baldwin#%&()  ?     
2           1005               Barbour 
3           1007             Bi#%&()bb 
4           1009         Blount  ?     
...          ...                    ...
3229       72151               Yabucoa 
3230       72153                Ya_uco 
3231       78010     ; St. Croix ?     
3232       78020               St. John
3233       78030             St. Thomas

[3234 rows x 2 columns]
<class 'pandas.core.frame.DataFrame'>

Escoger Filas

Podemos escoger ciertas filas a través de su posición. utilizando .iloc

from seaborn import load_dataset

print(counties.iloc[0:1,])
print(type(counties.iloc[0:1,]))

   codestate  codecounty         county  population     area
0          1        1001  Auta#%&()uga      54571.0  594.436
<class 'pandas.core.frame.DataFrame'>

print(counties.iloc[8:12,1:3])

    codecounty                 county
8         1017     ; Chambers  ?     
9         1019              Cherokee 
10        1021             ; Chilton 
11        1023               Choctaw 

O también escoger filas con base a alguna condicion o característica específica.

counties.tail()

	codestate	codecounty	county	population	area
3229	72	72151	Yabucoa	37941.0	55.215000
3230	72	72153	Ya_uco	42043.0	68.192000
3231	78	78010	; St. Croix ?	50601.0	83.345868
3232	78	78020	St. John	4170.0	19.689867
3233	78	78030	St. Thomas	51634.0	31.313503

counties["codestate"] > 72 #Verdadero o falso si cumple la condición

0       False
1       False
2       False
3       False
4       False
        ...  
3229    False
3230    False
3231     True
3232     True
3233     True
Name: codestate, Length: 3234, dtype: bool

codsat72 = counties[counties["codestate"] > 72] # Llama aquellos cuyo valor es verdadero
codsat72.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 3 entries, 3231 to 3233
Data columns (total 5 columns):
 #   Column      Non-Null Count  Dtype  
---  ------      --------------  -----  
 0   codestate   3 non-null      int64  
 1   codecounty  3 non-null      int64  
 2   county      3 non-null      object 
 3   population  3 non-null      float64
 4   area        3 non-null      float64
dtypes: float64(2), int64(2), object(1)
memory usage: 144.0+ bytes

Filas que no tengan valores faltantes en determinada columna

no_na = counties[counties["population"].notna()]

print(no_na.shape)
print(counties.shape)

(3232, 5)
(3234, 5)

Filas con valores específicos

Existen varias maneras de realizar estas búsquedas, algunas de ellas son utilizando isin, loc de la librería seaborn y or.

codcou1 = counties[counties["codecounty"].isin([78010, 72151])]
codcou1

	codestate	codecounty	county	population	area
3229	72	72151	Yabucoa	37941.0	55.215000
3231	78	78010	; St. Croix ?	50601.0	83.345868

codcou2 = counties.loc[:, 'codecounty'] == 78010
codcou2_ = counties.loc[codcou2]
codcou2_

	codestate	codecounty	county	population	area
3231	78	78010	; St. Croix ?	50601.0	83.345868

counties[(counties["codecounty"] == 78010) | (counties["codecounty"] == 72151)]

	codestate	codecounty	county	population	area
3229	72	72151	Yabucoa	37941.0	55.215000
3231	78	78010	; St. Croix ?	50601.0	83.345868

Filas con más de una característica específica.

counties[(counties["codestate"] == 72) & (counties["area"] >= 60)].info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 16 entries, 3153 to 3230
Data columns (total 5 columns):
 #   Column      Non-Null Count  Dtype  
---  ------      --------------  -----  
 0   codestate   16 non-null     int64  
 1   codecounty  16 non-null     int64  
 2   county      16 non-null     object 
 3   population  16 non-null     float64
 4   area        16 non-null     float64
dtypes: float64(2), int64(2), object(1)
memory usage: 768.0+ bytes

Eliminar valores omitidos

#Eliminar valores omitidos: dropna
# axis= 0 -> Eliminar fila completa
# axis=1 -> Eliminar columna completa
counties.dropna(subset=["population"], axis=0, inplace=False) #inplace:modifique directamente la BD

	codestate	codecounty	county	population	area
0	1	1001	Auta#%&()uga	54571.0	594.436000
1	1	1003	Baldwin#%&() ?	182265.0	1589.784000
2	1	1005	Barbour	27457.0	884.876000
3	1	1007	Bi#%&()bb	22915.0	622.582000
4	1	1009	Blount ?	57322.0	644.776000
...	...	...	...	...	...
3229	72	72151	Yabucoa	37941.0	55.215000
3230	72	72153	Ya_uco	42043.0	68.192000
3231	78	78010	; St. Croix ?	50601.0	83.345868
3232	78	78020	St. John	4170.0	19.689867
3233	78	78030	St. Thomas	51634.0	31.313503

3232 rows × 5 columns

Manipulación de datos textuales (Limpieza)

Para realizarle limpieza a columna county de la base de datos counties, podemos utilizar lo siguiente:

• **lower:**Poner en minúscula todo el texto.

• **replace:**Remplazar ciertos valores por otros.

• **strip:**Eliminar los espacios al principio y al final de la cadena.

• **title:**Poner primera letra de cada palabra en mayúscula.

counties["county"] = (counties["county"]
                            .str.lower()
                            .str.replace("[^a-záéíóúüñ ]","")
                            .str.replace(" +"," ")
                            .str.strip()
                            .str.title()
                           ) 

/opt/hostedtoolcache/Python/3.7.13/x64/lib/python3.7/site-packages/ipykernel_launcher.py:3: FutureWarning: The default value of regex will change from True to False in a future version.
  This is separate from the ipykernel package so we can avoid doing imports until
/opt/hostedtoolcache/Python/3.7.13/x64/lib/python3.7/site-packages/ipykernel_launcher.py:4: FutureWarning: The default value of regex will change from True to False in a future version.
  after removing the cwd from sys.path.

print(counties)

      codestate  codecounty     county  population         area
0             1        1001    Autauga     54571.0   594.436000
1             1        1003    Baldwin    182265.0  1589.784000
2             1        1005    Barbour     27457.0   884.876000
3             1        1007       Bibb     22915.0   622.582000
4             1        1009     Blount     57322.0   644.776000
...         ...         ...        ...         ...          ...
3229         72       72151    Yabucoa     37941.0    55.215000
3230         72       72153      Yauco     42043.0    68.192000
3231         78       78010   St Croix     50601.0    83.345868
3232         78       78020    St John      4170.0    19.689867
3233         78       78030  St Thomas     51634.0    31.313503

[3234 rows x 5 columns]

Creación de columnas a partir de otras

counties["densidad"]= counties["population"]/counties["area"]
print(counties[["densidad","population","area"]])

         densidad  population         area
0       91.802986     54571.0   594.436000
1      114.647650    182265.0  1589.784000
2       31.029206     27457.0   884.876000
3       36.806397     22915.0   622.582000
4       88.902192     57322.0   644.776000
...           ...         ...          ...
3229   687.150231     37941.0    55.215000
3230   616.538597     42043.0    68.192000
3231   607.120676     50601.0    83.345868
3232   211.784062      4170.0    19.689867
3233  1648.937201     51634.0    31.313503

[3234 rows x 3 columns]

Renombrar columnas

print(counties.columns)
counties.rename(columns={"densidad":"densidad_pob"},inplace=True) #renombrar columna
print(counties.columns) #Verificación

Index(['codestate', 'codecounty', 'county', 'population', 'area', 'densidad'], dtype='object')
Index(['codestate', 'codecounty', 'county', 'population', 'area',
       'densidad_pob'],
      dtype='object')

Modificar tipo de dato de una columna

print(counties.info())
print("\n")
counties["codestate"]=counties["codestate"].astype("float")
print(counties.info())

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3234 entries, 0 to 3233
Data columns (total 6 columns):
 #   Column        Non-Null Count  Dtype  
---  ------        --------------  -----  
 0   codestate     3234 non-null   int64  
 1   codecounty    3234 non-null   int64  
 2   county        3234 non-null   object 
 3   population    3232 non-null   float64
 4   area          3234 non-null   float64
 5   densidad_pob  3232 non-null   float64
dtypes: float64(3), int64(2), object(1)
memory usage: 151.7+ KB
None


<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3234 entries, 0 to 3233
Data columns (total 6 columns):
 #   Column        Non-Null Count  Dtype  
---  ------        --------------  -----  
 0   codestate     3234 non-null   float64
 1   codecounty    3234 non-null   int64  
 2   county        3234 non-null   object 
 3   population    3232 non-null   float64
 4   area          3234 non-null   float64
 5   densidad_pob  3232 non-null   float64
dtypes: float64(4), int64(1), object(1)
memory usage: 151.7+ KB
None

Eliminar filas y columnas

Eliminación filas, puede ser por posición o que cumpla una característica.

print(counties.shape)
counties = counties.drop(counties.iloc[0:3,].index)
counties = counties.drop(counties[counties['codecounty']==70].index)
print(counties.shape)

(3234, 6)
(3231, 6)

counties.drop(['densidad_pob'], axis=1,inplace=True) #inplace=True, para que me modifique la BD
counties.columns

Index(['codestate', 'codecounty', 'county', 'population', 'area'], dtype='object')

Conteos

La función groupby devuelve un objeto DataFrameGroupBy que contiene la información agrupada según las especificaciones que indiquemos.

group=counties.groupby("codestate")
print(type(group))

<class 'pandas.core.groupby.generic.DataFrameGroupBy'>

Podemos obtener detalles del anterior dataframe

print(group.first().head(10)) # Nos muestra el primer elemento de cada grupo según "codestate"
print('\n')
print(group.get_group(78)) #Nos muestra los elementos cuyo codestate es 78

           codecounty                county  population       area
codestate                                                         
1.0              1007                  Bibb     22915.0    622.582
2.0              2013        Aleutians East      3141.0   6981.943
4.0              4001                Apache     71518.0  11197.521
5.0              5001              Arkansas     19019.0    988.770
6.0              6001               Alameda   1510271.0    739.017
8.0              8001                 Adams    441603.0   1167.653
9.0              9001             Fairfield    916829.0    624.890
10.0            10001                  Kent    162310.0    586.179
11.0            11001  District Of Columbia    601723.0     61.048
12.0            12001               Alachua    247336.0    875.020


      codestate  codecounty     county  population       area
3231       78.0       78010   St Croix     50601.0  83.345868
3232       78.0       78020    St John      4170.0  19.689867
3233       78.0       78030  St Thomas     51634.0  31.313503

counties.groupby("codestate")["codecounty"].count().head(10)

codestate
1.0     64
2.0     29
4.0     15
5.0     75
6.0     58
8.0     64
9.0      8
10.0     3
11.0     1
12.0    67
Name: codecounty, dtype: int64

Modificar Tablas

es posible hacer cambios de orden y estructura en las tablas de pandas.

folder_path="https://github.com/AprendizajeProfundo/Libro_Fundamentos_Programacion/blob/main/Python/Datos/"

elections=pd.read_excel(folder_path+'elections.xlsx?raw=true')
elections.head(5)

	year	democrat	republic	other	codecounty
0	2000	4942	11993	273	1001
1	2000	13997	40872	1611	1003
2	2000	5188	5096	111	1005
3	2000	2710	4273	118	1007
4	2000	4977	12667	329	1009

organicemos los datos descendentemente depende de la cantidad de votos democratas

elections.sort_values(["democrat"], ascending=False)

	year	democrat	republic	other	codecounty
12626	2016	2464364	769743	200201	6037
6400	2008	2295853	956425	65970	6037
9513	2012	2216903	885333	78831	6037
3287	2004	1907736	1076225	37686	6037
175	2000	1710505	871930	112719	6037
...	...	...	...	...	...
11981	2012	9	54	1	48301
8852	2008	8	151	4	48269
11965	2012	5	139	1	48269
15078	2016	5	149	5	48269
15094	2016	4	58	3	48301

15563 rows × 5 columns

usando groupby() podemos juntar los datos por año y sumarlos para tener el total de votos por año

elections.groupby('year')[['democrat', 'republic']].sum()

	democrat	republic
year
2000	50922964	50290498
2004	58614073	61603653
2008	69204582	59703000
2012	65508961	60460609
2016	65626398	62790289

Podemos aplicar ambos métodos en un mismo código

elections_sort=(elections.groupby('year')[['democrat', 'republic']]
                .sum()
                .sort_values('democrat', ascending=False))
display(elections_sort)

	democrat	republic
year
2008	69204582	59703000
2016	65626398	62790289
2012	65508961	60460609
2004	58614073	61603653
2000	50922964	50290498

Para reestructurar los datos usamos la función pivot().

pivot_elections=elections.pivot(index='codecounty',
                columns='year',
                values=['democrat','republic'])
pivot_elections.head()

	democrat					republic
year	2000	2004	2008	2012	2016	2000	2004	2008	2012	2016
codecounty
1001	4942.0	4758.0	6093.0	6363.0	5936.0	11993.0	15196.0	17403.0	17379.0	18172.0
1003	13997.0	15599.0	19386.0	18424.0	18458.0	40872.0	52971.0	61271.0	66016.0	72883.0
1005	5188.0	4832.0	5697.0	5912.0	4871.0	5096.0	5899.0	5866.0	5550.0	5454.0
1007	2710.0	2089.0	2299.0	2202.0	1874.0	4273.0	5472.0	6262.0	6132.0	6738.0
1009	4977.0	3938.0	3522.0	2970.0	2156.0	12667.0	17386.0	20389.0	20757.0	22859.0

Múltiples índices

Se pueden elegir múltiples variables como índice del dataframe. Esto es util para facilitar la extracción de información en ciertos casos

counties_multi=pd.read_excel(folder_path+"counties.xlsx?raw=true", index_col=[0, 1])
display(counties_multi)

		county	population	area
codestate	codecounty
1	1001	Auta#%&()uga	54571.0	594.436000
	1003	Baldwin#%&() ?	182265.0	1589.784000
	1005	Barbour	27457.0	884.876000
	1007	Bi#%&()bb	22915.0	622.582000
	1009	Blount ?	57322.0	644.776000
...	...	...	...	...
72	72151	Yabucoa	37941.0	55.215000
	72153	Ya_uco	42043.0	68.192000
78	78010	; St. Croix ?	50601.0	83.345868
	78020	St. John	4170.0	19.689867
	78030	St. Thomas	51634.0	31.313503

3234 rows × 3 columns

Podemos obtener la suma de población de cada estado

counties_multi['population'].sum(level='codestate').head(5)

/opt/hostedtoolcache/Python/3.7.13/x64/lib/python3.7/site-packages/ipykernel_launcher.py:1: FutureWarning: Using the level keyword in DataFrame and Series aggregations is deprecated and will be removed in a future version. Use groupby instead. df.sum(level=1) should use df.groupby(level=1).sum().
  """Entry point for launching an IPython kernel.

codestate
1     4779736.0
2      710231.0
4     6392017.0
5     2915918.0
6    37253956.0
Name: population, dtype: float64

Concatenar y unir

Es posible unir varias tablas tanto vertical como horizontalmente.

En ambos casos, podemos usar concat(), y se juntarán las bases con bases a los nombres de columnas o los indices de las filas

elections_2000=elections[elections['year']==2000]
display(elections_2000) #display:ejecuta el método dunder apropiado para obtener los datos apropiados para mostrar

	year	democrat	republic	other	codecounty
0	2000	4942	11993	273	1001
1	2000	13997	40872	1611	1003
2	2000	5188	5096	111	1005
3	2000	2710	4273	118	1007
4	2000	4977	12667	329	1009
...	...	...	...	...	...
3107	2000	5521	9425	745	56037
3108	2000	4019	5454	958	56039
3109	2000	1650	5469	299	56041
3110	2000	806	3138	107	56043
3111	2000	449	2521	90	56045

3112 rows × 5 columns

elections_2004=elections[elections['year']==2004]
elections_2004['dummy']=0
display(elections_2004)

/opt/hostedtoolcache/Python/3.7.13/x64/lib/python3.7/site-packages/ipykernel_launcher.py:2: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
  

	year	democrat	republic	other	codecounty	dummy
3112	2004	4758	15196	127	1001	0
3113	2004	15599	52971	750	1003	0
3114	2004	4832	5899	46	1005	0
3115	2004	2089	5472	39	1007	0
3116	2004	3938	17386	180	1009	0
...	...	...	...	...	...	...
6220	2004	5208	10653	411	56037	0
6221	2004	5972	5124	263	56039	0
6222	2004	1815	6081	185	56041	0
6223	2004	855	3200	59	56043	0
6224	2004	580	2739	73	56045	0

3113 rows × 6 columns

Si se hace la concatenación sin más, se tomarán todas las columnas y se agregarán NaN

elections_00_04=pd.concat([elections_2000, elections_2004])
display(elections_00_04)

	year	democrat	republic	other	codecounty	dummy
0	2000	4942	11993	273	1001	NaN
1	2000	13997	40872	1611	1003	NaN
2	2000	5188	5096	111	1005	NaN
3	2000	2710	4273	118	1007	NaN
4	2000	4977	12667	329	1009	NaN
...	...	...	...	...	...	...
6220	2004	5208	10653	411	56037	0.0
6221	2004	5972	5124	263	56039	0.0
6222	2004	1815	6081	185	56041	0.0
6223	2004	855	3200	59	56043	0.0
6224	2004	580	2739	73	56045	0.0

6225 rows × 6 columns

Ahora, usando merge(), podemos hacer uniones de las tablas de forma horizontal que compartan una columna/indice en común

Tipos de Joins

display(counties)

	codestate	codecounty	county	population	area
3	1.0	1007	Bibb	22915.0	622.582000
4	1.0	1009	Blount	57322.0	644.776000
5	1.0	1011	Bullock	10914.0	622.805000
6	1.0	1013	Butler	20947.0	776.829000
7	1.0	1015	Calhoun	118572.0	605.868000
...	...	...	...	...	...
3229	72.0	72151	Yabucoa	37941.0	55.215000
3230	72.0	72153	Yauco	42043.0	68.192000
3231	78.0	78010	St Croix	50601.0	83.345868
3232	78.0	78020	St John	4170.0	19.689867
3233	78.0	78030	St Thomas	51634.0	31.313503

3231 rows × 5 columns

inner_joined=pd.merge(elections, counties)
inner_joined.head(10)

	year	democrat	republic	other	codecounty	codestate	county	population	area
0	2000	2710	4273	118	1007	1.0	Bibb	22915.0	622.582
1	2004	2089	5472	39	1007	1.0	Bibb	22915.0	622.582
2	2008	2299	6262	83	1007	1.0	Bibb	22915.0	622.582
3	2012	2202	6132	86	1007	1.0	Bibb	22915.0	622.582
4	2016	1874	6738	207	1007	1.0	Bibb	22915.0	622.582
5	2000	4977	12667	329	1009	1.0	Blount	57322.0	644.776
6	2004	3938	17386	180	1009	1.0	Blount	57322.0	644.776
7	2008	3522	20389	356	1009	1.0	Blount	57322.0	644.776
8	2012	2970	20757	279	1009	1.0	Blount	57322.0	644.776
9	2016	2156	22859	573	1009	1.0	Blount	57322.0	644.776

outer_joined=pd.merge(elections, counties, how='outer')
outer_joined.head(10)

	year	democrat	republic	other	codecounty	codestate	county	population	area
0	2000.0	4942.0	11993.0	273.0	1001	NaN	NaN	NaN	NaN
1	2004.0	4758.0	15196.0	127.0	1001	NaN	NaN	NaN	NaN
2	2008.0	6093.0	17403.0	145.0	1001	NaN	NaN	NaN	NaN
3	2012.0	6363.0	17379.0	190.0	1001	NaN	NaN	NaN	NaN
4	2016.0	5936.0	18172.0	865.0	1001	NaN	NaN	NaN	NaN
5	2000.0	13997.0	40872.0	1611.0	1003	NaN	NaN	NaN	NaN
6	2004.0	15599.0	52971.0	750.0	1003	NaN	NaN	NaN	NaN
7	2008.0	19386.0	61271.0	756.0	1003	NaN	NaN	NaN	NaN
8	2012.0	18424.0	66016.0	898.0	1003	NaN	NaN	NaN	NaN
9	2016.0	18458.0	72883.0	3874.0	1003	NaN	NaN	NaN	NaN

left_joined=pd.merge(elections, counties, how='left')
left_joined.head(10)

	year	democrat	republic	other	codecounty	codestate	county	population	area
0	2000	4942	11993	273	1001	NaN	NaN	NaN	NaN
1	2000	13997	40872	1611	1003	NaN	NaN	NaN	NaN
2	2000	5188	5096	111	1005	NaN	NaN	NaN	NaN
3	2000	2710	4273	118	1007	1.0	Bibb	22915.0	622.582
4	2000	4977	12667	329	1009	1.0	Blount	57322.0	644.776
5	2000	3395	1433	76	1011	1.0	Bullock	10914.0	622.805
6	2000	3606	4127	70	1013	1.0	Butler	20947.0	776.829
7	2000	15781	22306	822	1015	1.0	Calhoun	118572.0	605.868
8	2000	5616	6037	181	1017	1.0	Chambers	34215.0	596.531
9	2000	3497	4154	172	1019	1.0	Cherokee	25989.0	553.700

right_joined=pd.merge(elections, counties, how='right')
right_joined.head(10)

	year	democrat	republic	other	codecounty	codestate	county	population	area
0	2000.0	2710.0	4273.0	118.0	1007	1.0	Bibb	22915.0	622.582
1	2004.0	2089.0	5472.0	39.0	1007	1.0	Bibb	22915.0	622.582
2	2008.0	2299.0	6262.0	83.0	1007	1.0	Bibb	22915.0	622.582
3	2012.0	2202.0	6132.0	86.0	1007	1.0	Bibb	22915.0	622.582
4	2016.0	1874.0	6738.0	207.0	1007	1.0	Bibb	22915.0	622.582
5	2000.0	4977.0	12667.0	329.0	1009	1.0	Blount	57322.0	644.776
6	2004.0	3938.0	17386.0	180.0	1009	1.0	Blount	57322.0	644.776
7	2008.0	3522.0	20389.0	356.0	1009	1.0	Blount	57322.0	644.776
8	2012.0	2970.0	20757.0	279.0	1009	1.0	Blount	57322.0	644.776
9	2016.0	2156.0	22859.0	573.0	1009	1.0	Blount	57322.0	644.776

Series de tiempo y time stamps

Existe un tipo de datos llamado “timestamp” que se usa para medir registros de tiempo. Es posible pasar strings a este tipo de dato en diferentes formatos y trabajar con este usándolo como indice

airrpm=pd.read_csv(folder_path+'airrpm.txt?raw=true', header=None, delimiter= '\s+', decimal=",")
airrpm.columns=["Time", "R1", "R2", "R3"]
airrpm.head()

	Time	R1	R2	R3
0	Jan-79	26.64	15.50	11.15
1	Feb-79	27.20	16.58	10.62
2	Mar-79	27.87	18.85	9.02
3	Apr-79	23.22	17.23	5.99
4	May-79	23.27	16.04	7.23

La variable “Time” en este momento es un string. Pero podemos transformarlo usando to_datetime(). Para obtener más información de qué se podría poner en format, pueden revisar este link

airrpm['Time']=pd.to_datetime(airrpm['Time'], format='%b-%y')
airrpm.index=airrpm['Time']
airrpm=airrpm.drop(['Time'], axis=1)
airrpm.head()

	R1	R2	R3
Time
1979-01-01	26.64	15.50	11.15
1979-02-01	27.20	16.58	10.62
1979-03-01	27.87	18.85	9.02
1979-04-01	23.22	17.23	5.99
1979-05-01	23.27	16.04	7.23

Definiendo la fecha como índice, podemos extraer información con respecto al año y respecto al mes

airrpm.groupby([airrpm.index.month])['R1'].describe()

	count	mean	std	min	25%	50%	75%	max
Time
1	26.0	43.736538	9.778402	26.64	36.6575	44.82	51.86	59.36
2	25.0	43.680800	9.983696	27.20	35.5400	44.74	52.73	59.74
3	25.0	44.514400	10.363336	27.87	35.7000	45.01	53.26	60.10
4	25.0	44.287200	10.480331	23.22	36.6100	45.63	52.79	60.49
5	25.0	43.891600	10.524814	23.27	34.6000	45.11	52.57	60.36
6	25.0	45.655600	10.586951	27.30	35.4200	47.29	54.55	61.88
7	25.0	46.478400	10.333910	29.28	38.2700	47.79	54.94	63.02
8	25.0	46.773600	10.613169	27.27	38.8600	48.06	55.39	63.71
9	25.0	44.202400	9.731246	27.13	37.2600	46.39	51.83	59.54
10	25.0	44.433200	10.093832	26.97	37.8500	45.89	51.56	60.26
11	25.0	43.841200	9.923719	26.19	37.3300	45.50	51.02	59.21
12	25.0	44.600400	9.573277	27.32	38.7500	45.33	51.98	57.65

airrpm.groupby([airrpm.index.year]).sum().head(10)

	R1	R2	R3
Time
1979	328.83	210.49	118.34
1980	340.95	200.92	140.05
1981	341.96	198.53	143.45
1982	355.17	210.23	144.91
1983	374.03	226.76	147.26
1984	415.88	243.71	172.18
1985	440.43	271.80	168.61
1986	492.37	302.20	190.18
1987	522.07	324.58	197.52
1988	530.25	328.69	201.57