In [1]:

%matplotlib inline
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns
pd.set_option('display.width', 500)
pd.set_option('display.max_columns', 100)

In [45]:

1+1

Out[45]:

here is a heading¶

a smaller one¶

an even smaller one¶

even less important¶

and here is some text

hello world

http://www.google.com

hello world

here is an equation $E = m c^2$

$\alpha = \beta$

hello bold, nor emph

In [11]:

#also https://dl.dropboxusercontent.com/u/75194/tao-full.csv
dfsmall=pd.read_csv("https://dl.dropboxusercontent.com/u/75194/tao.csv")
dfsmall.head()

Out[11]:

	Unnamed: 0	year	longitude	sea.surface.temp	air.temp	humidity	uwind	vwind
0	1	1997	-110	27.590000	27.15	79.599998	-6.4	5.4
1	2	1997	-110	27.549999	27.02	75.800003	-5.3	5.3
2	3	1997	-110	27.570000	27.00	76.500000	-5.1	4.5
3	4	1997	-110	27.620001	26.93	76.199997	-4.9	2.5
4	5	1997	-110	27.650000	26.84	76.400002	-3.5	4.1

Primary question: Can we detect an El Nino event, based on sea surface temperature? What about other observed variables?

In [13]:

del dfsmall["Unnamed: 0"]
dfsmall.head()

Out[13]:

	year	longitude	sea.surface.temp	air.temp	humidity	uwind	vwind
0	1997	-110	27.590000	27.15	79.599998	-6.4	5.4
1	1997	-110	27.549999	27.02	75.800003	-5.3	5.3
2	1997	-110	27.570000	27.00	76.500000	-5.1	4.5
3	1997	-110	27.620001	26.93	76.199997	-4.9	2.5
4	1997	-110	27.650000	26.84	76.400002	-3.5	4.1

In [12]:

dfsmall.dtypes

Out[12]:

Unnamed: 0            int64
year                  int64
latitude              int64
longitude             int64
sea.surface.temp    float64
air.temp            float64
humidity            float64
uwind               float64
vwind               float64
dtype: object

In [14]:

dfsmall.shape

Out[14]:

(736, 8)

In [16]:

dfsmall.groupby('year').mean()#look up groupby

Out[16]:

	latitude	longitude	sea.surface.temp	air.temp	humidity	uwind	vwind
year
1993	-1.00	-102.5	23.542110	23.365962	86.847636	-3.828804	2.496196
1997	-1.75	-102.5	28.168859	27.109794	82.632065	-3.602989	2.775272

In [39]:

dfsmall.groupby('year').describe()

Out[39]:

		air.temp	humidity	latitude	longitude	sea.surface.temp	uwind	vwind
year
1993	count	364.000000	275.000000	368.000000	368.000000	365.000000	368.000000	368.000000
	mean	23.365962	86.847636	-1.000000	-102.500000	23.542110	-3.828804	2.496196
	std	0.810965	2.936683	1.001361	7.510211	0.877629	1.819194	1.595203
	min	21.420000	79.800003	-2.000000	-110.000000	21.600000	-7.800000	-4.300000
	25%	22.730000	85.149998	-2.000000	-110.000000	22.840000	-5.100000	1.600000
	50%	23.379999	87.400002	-1.000000	-102.500000	23.500000	-3.900000	2.600000
	75%	23.900000	89.000000	0.000000	-95.000000	24.070000	-2.700000	3.500000
	max	25.030001	92.199997	0.000000	-95.000000	26.379999	3.300000	6.300000
1997	count	291.000000	368.000000	368.000000	368.000000	368.000000	368.000000	368.000000
	mean	27.109794	82.632065	-1.750000	-102.500000	28.168859	-3.602989	2.775272
	std	0.617771	4.773291	2.049124	7.510211	0.710235	2.085480	2.516228
	min	25.200001	71.599998	-5.000000	-110.000000	26.480000	-8.100000	-6.200000
	25%	26.790001	79.074999	-2.750000	-110.000000	27.627500	-5.100000	1.375000
	50%	27.150000	82.500000	-1.000000	-102.500000	28.209999	-3.800000	3.500000
	75%	27.575000	86.500000	0.000000	-95.000000	28.662500	-2.300000	4.600000
	max	28.500000	94.800003	0.000000	-95.000000	30.170000	4.300000	7.300000

In [19]:

dfsmall.columns

Out[19]:

Index([u'year', u'latitude', u'longitude', u'sea.surface.temp', u'air.temp', u'humidity', u'uwind', u'vwind'], dtype='object')

In [20]:

for feat in dfsmall.columns:
    print feat, dfsmall[feat].isnull().mean()

year 0.0
latitude 0.0
longitude 0.0
sea.surface.temp 0.00407608695652
air.temp 0.110054347826
humidity 0.126358695652
uwind 0.0
vwind 0.0

In [27]:

for k,v in dfsmall.groupby('year'):
    for feat in dfsmall.columns:
        print k, feat, v[feat].isnull().mean()

1993 year 0.0
1993 latitude 0.0
1993 longitude 0.0
1993 sea.surface.temp 0.00815217391304
1993 air.temp 0.0108695652174
1993 humidity 0.252717391304
1993 uwind 0.0
1993 vwind 0.0
1997 year 0.0
1997 latitude 0.0
1997 longitude 0.0
1997 sea.surface.temp 0.0
1997 air.temp 0.209239130435
1997 humidity 0.0
1997 uwind 0.0
1997 vwind 0.0

In [28]:

dfsmallnew=dfsmall[dfsmall['humidity'].notnull()]

In [29]:

dfsmallnew.shape

Out[29]:

(643, 8)

In [30]:

dfsmallnew=dfsmallnew[dfsmallnew['air.temp'].notnull()]
dfsmallnew=dfsmallnew[dfsmallnew['sea.surface.temp'].notnull()]
dfsmallnew.shape

Out[30]:

(565, 8)

In [31]:

dfsmallnew.head()

Out[31]:

	year	longitude	sea.surface.temp	air.temp	humidity	uwind	vwind
0	1997	-110	27.590000	27.15	79.599998	-6.4	5.4
1	1997	-110	27.549999	27.02	75.800003	-5.3	5.3
2	1997	-110	27.570000	27.00	76.500000	-5.1	4.5
3	1997	-110	27.620001	26.93	76.199997	-4.9	2.5
4	1997	-110	27.650000	26.84	76.400002	-3.5	4.1

In [32]:

dfsmallnew.groupby('year').count()

Out[32]:

	latitude	longitude	sea.surface.temp	air.temp	humidity	uwind	vwind
year
1993	274	274	274	274	274	274	274
1997	291	291	291	291	291	291	291

In [34]:

sns.pairplot(dfsmallnew, hue="year", size=2.5, alpha=0.3);#ahap happened here?
#look up http://stanford.edu/~mwaskom/software/seaborn/tutorial/axis_grids.html

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-34-059f20c0f4f1> in <module>()
----> 1 sns.pairplot(dfsmallnew, hue="year", size=2.5, alpha=0.3);

TypeError: pairplot() got an unexpected keyword argument 'alpha'

In [35]:

sns.pairplot(dfsmallnew, hue="year", size=2.5, plot_kws={'alpha':0.4});

In [36]:

vars=list(dfsmallnew.columns)
vars

Out[36]:

['year',
 'latitude',
 'longitude',
 'sea.surface.temp',
 'air.temp',
 'humidity',
 'uwind',
 'vwind']

In [38]:

sns.pairplot(dfsmallnew, hue="year", size=2.5, plot_kws={'alpha':0.4}, vars=vars[1:]);

In [40]:

g = sns.PairGrid(dfsmallnew, vars=["sea.surface.temp", "air.temp"], hue="year")
g.map(plt.scatter)

In [43]:

g = sns.PairGrid(dfsmallnew, vars=["sea.surface.temp", "air.temp"], hue="year")
g.map_upper(plt.scatter, alpha=0.3)
g.map_diag(sns.kdeplot, alpha=0.5)
g.map_lower(sns.kdeplot, alpha=0.5)

In [ ]:

dfsmall.