spreg.MoranRes¶

class spreg.MoranRes(ols, w, z=False)[source]¶

Moran’s I for spatial autocorrelation in residuals from OLS regression

Parameters:

olsOLS: OLS regression object
wW: Spatial weights instance
zbool: If set to True computes attributes eI, vI and zI. Due to computational burden of vI, defaults to False.

Attributes:

Ifloat: Moran’s I statistic
eIfloat: Moran’s I expectation
vIfloat: Moran’s I variance
zIfloat: Moran’s I standardized value

Examples

>>> import numpy as np
>>> import libpysal
>>> from spreg import OLS
>>> import spreg

Open the csv file to access the data for analysis

>>> csv = libpysal.io.open(libpysal.examples.get_path('columbus.dbf'),'r')

Pull out from the csv the files we need (‘HOVAL’ as dependent as well as ‘INC’ and ‘CRIME’ as independent) and directly transform them into nx1 and nx2 arrays, respectively

>>> y = np.array([csv.by_col('HOVAL')]).T
>>> x = np.array([csv.by_col('INC'), csv.by_col('CRIME')]).T

Create the weights object from existing .gal file

>>> w = libpysal.io.open(libpysal.examples.get_path('columbus.gal'), 'r').read()

Row-standardize the weight object (not required although desirable in some cases)

>>> w.transform='r'

Run an OLS regression

>>> ols = OLS(y, x)

Run Moran’s I test for residual spatial autocorrelation in an OLS model. This computes the traditional statistic applying a correction in the expectation and variance to account for the fact it comes from residuals instead of an independent variable

>>> m = spreg.MoranRes(ols, w, z=True)

Value of the Moran’s I statistic:

>>> print(round(m.I,4))
0.1713

Value of the Moran’s I expectation:

>>> print(round(m.eI,4))
-0.0345

Value of the Moran’s I variance:

>>> print(round(m.vI,4))
0.0081

Value of the Moran’s I standardized value. This is distributed as a standard Normal(0, 1)

>>> print(round(m.zI,4))
2.2827

P-value of the standardized Moran’s I value (z):

>>> print(round(m.p_norm,4))
0.0224

__init__(ols, w, z=False)[source]¶

Methods

__init__(ols, w[, z])