In [1]:
from pylab import *
import pandas as pd
import pandas.tools.plotting as pt
import datetime as dt
from pandas.stats import moments as mm
from sklearn import linear_model as lin
In [2]:
nino34=pd.read_csv("http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/detrend.nino34.ascii.txt", sep="\s+")
link = pd.read_csv("https://raw.githubusercontent.com/johncarlosbaez/el-nino/master/R/average-link-strength-1948-2013.txt",
sep="\s+", header=None, names=["N", "S"])
link.S-=link.S.mean()
dates = dt.date(1948,01,01)+vectorize(dt.timedelta)(days=10*link["N"]+729)
del link["N"]
#dates
In [3]:
link.shape
Out[3]:
In [4]:
nino34.ANOM.plot()
Out[4]:
In [5]:
link.plot()
Out[5]:
In [6]:
zz=specgram(link.T.values[-1])
In [7]:
zz=psd(link.T.values[-1], NFFT=2**9, noverlap=2**8)
In [8]:
figure(figsize=(5,5))
plot(log2(zz[1][1:]),log2(zz[0][1:]))
Out[8]:
In [9]:
nino34.T
Out[9]:
In [10]:
linktmp=link[arange(len(link))%73 != 0]
ll= len(linktmp)
print ll
linkmnth= linktmp.groupby(arange(ll)//3).mean()
linkmnth.reset_index()
linkmnth.T
linkmnth.shape
Out[10]:
In [11]:
linkx=linkmnth
df=nino34.iloc[:len(linkx)]
df.index=range(df.shape[0])
print ll, len(linkx), df.shape
df["link"]=linkmnth
In [12]:
df.to_csv("ANOMALY.csv")
df.T
Out[12]:
Link vs ANOM, no lag
In [13]:
df.plot(x="link", y="ANOM", kind="scatter", figsize=(9,9), alpha=0.3, marker="o")
Out[13]:
Link vs ANOM, 6 month lag
In [14]:
figure(figsize=(9,9))
scatter(df["link"].iloc[:-6], y=df["ANOM"].iloc[6:], alpha=0.3, marker="o")
Out[14]:
ANOM vs ANOM, 6 month lag
In [15]:
figure(figsize=(9,9))
scatter(df.ANOM[:-6], y=df.ANOM[6:], alpha=0.3, marker="o")
Out[15]:
ANOM & link vs ANOM, 6 month lag
In [16]:
reg = lin.LinearRegression()
reg.fit(df[["ANOM","link"]].iloc[:-6], df["ANOM"].iloc[6:])
print reg.coef_, reg.intercept_
print reg.score(df[["ANOM","link"]].iloc[:-6], df["ANOM"].iloc[6:])
In [17]:
figure(figsize=(9,9))
scatter(reg.predict(df[["ANOM","link"]].iloc[:-6]),
df["ANOM"].iloc[6:],
alpha=0.3, marker="o")
Out[17]:
ANOM auto courrelation
In [18]:
figure(figsize=(20,3))
zz=acorr(df.ANOM, maxlags=None)
In [19]:
figure(figsize=(20,3))
zz=acorr(df.ANOM, maxlags=48)
Link autocorrelation
In [20]:
figure(figsize=(20,3))
zz=acorr(df.link, maxlags=None )
In [21]:
figure(figsize=(20,3))
zz=acorr(df.link, maxlags=48 )
Link-ANOM cross-correlation
In [22]:
figure(figsize=(20,3))
zz=xcorr(df.link, df.ANOM, maxlags=None)
In [23]:
mx= argmin(zz[1]), argmax(zz[1])
print mx, zz[0][mx[0]],zz[1][mx[0]], zz[0][mx[1]], zz[1][mx[1]]
In [24]:
figure(figsize=(20,3))
zz=xcorr(df.link, df.ANOM, maxlags=48)
In [25]:
mx= argmin(zz[1]), argmax(zz[1])
print mx, zz[0][mx[0]],zz[1][mx[0]], zz[0][mx[1]], zz[1][mx[1]]
In [26]:
print zz[1][zz[0]==0]
Ignore below¶
In [27]:
oni = mm.rolling_mean(nino34.ANOM.values, 3)[2:]
elnino = (mm.rolling_min(oni, 5)[4:] - 0.5)
ludnino = (mm.rolling_min(nino34.ANOM, 5)[4:] - 0.5)
oni -= oni.mean()
elnino -= elnino.mean()
ludnino -= ludnino.mean()
In [28]:
zz=acorr(elnino,maxlags=None)
In [29]:
zz=acorr(ludnino,maxlags=None)
In [30]:
zz=xcorr(ludnino,maxlags=None)
In []:
aa=pd.DataFrame({"link":linkmnth["S"], "elnino": elnino[:linkmnth.shape[0]]})
aa.plot()#, maxlags=None)
In []:
zz=pt.scatter_matrix(df, alpha=0.2, diagonal="kde", figsize=(15,15))
In []:
figure(figsize=(20,3))
zz=cohere(df.ANOM, df.link, NFFT=64, noverlap=32)
In []:
xcorr(aa.link, aa.elnino,maxlags=48)
In []:
linkmnth.shape
In []:
figure(figsize=(10,10))
plot(elnino[:-6], elnino[6:],marker=".", alpha=0.2)
In []:
plot(linkmnth["S"], elnino[:linkmnth.shape[0]], marker=".", alpha=0.2)
In []:
figure(figsize=(10,10))
plot(linkmnth["S"], ludnino[2:2+linkmnth.shape[0]], marker=".", alpha=0.2)
In []:
scatter(nino34.ANOM[6:], elnino)
In []:
z=arange(-10,10,0.1)
k=20
df=pd.DataFrame({"a": sin(z[k:]), "b":sin(z[:-k])})
df.plot("a","b")
In []:
df.plot("a","b",kind="scatter")
In []:
zz=xcorr(df.a, df.b, maxlags=40)
In []: