def reg_3():
# SOURCE: Hansen, Chapters 4.19, page 126.
filename = data_path('cps09mar.txt')
frame = pd.read_csv(filename)
frame = featurize(frame)
frame = filter_3(frame)
y = 'log(wage)'
x = [
'education',
'experience',
'experience^2/100',
'female',
'female*union',
'male*union',
'female*married',
'male*married',
'female*fromerly_married',
'male*formerly_married',
'hispanic',
'black',
'american_indian',
'asian',
'mixed_race',
'intercept',
]
reg = regress.Reg(x, y)
reg.fit(frame)
# Chapter 4.19, pages 126-127
reg.summarize() # OUT: table 4.2, page 127
def reg_1():
# SOURCE: Hansen, Chapters 4.21, equation 3.4.41, page 130
filename = data_path('ddk2011.txt')
frame = pd.read_csv(filename)
frame = featurize(frame)
frame = filter_1(frame)
x = ['intercept', 'tracking']
y = 'testscore'
reg = regress.Reg(x, y)
reg.fit(frame)
reg.summarize()
def reg_2():
# SOURCE: Hansen, Chapters 4.21, equation 3.4.41, page 130
filename = data_path('ddk2011.txt')
frame = pd.read_csv(filename)
frame = featurize(frame)
frame = filter_1(frame)
x = ['intercept', 'tracking']
y = 'testscore'
grp = 'schoolid'
reg = regress.Cluster(x, y, grp)
reg.fit(frame)
reg.summarize(vce='cr3') # OUT: see Hansen's equation 4.52, page 134
def reg_2():
# SOURCE: Hansen, Chapters 3.7, equation 3.14, pages 76.
filename = data_path('cps09mar.txt')
frame = pd.read_csv(filename)
frame = featurize(frame)
frame = filter_2(frame)
y = 'log(wage)'
x = ['intercept', 'education', 'experience', 'experience^2']
reg = regress.Reg(x, y)
reg.fit(frame)
# Chapter 3.21, page 92
print(frame.shape) # OUT: (268, 18)
print(reg.influence()) # OUT: 29%
def reg_1():
# SOURCE: Hansen, Chapters 3.7, equation 3.13, page 75
filename = data_path('cps09mar.txt')
frame = pd.read_csv(filename)
frame = featurize(frame)
frame = filter_1(frame)
y = 'log(wage)'
x = ['education', 'intercept']
reg = regress.Reg(x, y)
reg.fit(frame)
# SOURCE: chapter 4.15, page 121
print(reg.vce(estimator='0')) # OUT: [[0.002, -0.031], [-0.031, 0.499]]
print(reg.vce(estimator='hc2')
) # OUT: [[ 0.0009314 -0.01479734], [-0.01479722 0.24282344]]
# SOURCE: chapter 4.15, page 122
print(reg.std_err(estimator='0')) # OUT: [0.045, 0.707]
print(reg.std_err(estimator='hc0')) # OUT: [0.029, 0.461]
print(reg.std_err(estimator='hc1')) # OUT: [0.030, 0.486]
print(reg.std_err(estimator='hc2')) # OUT: [0.031, 0.493]
print(reg.std_err(estimator='hc3')) # OUT: [0.033, 0.527]
def reg_4():
# see Chapter 7.11
filename = data_path('cps09mar.txt')
frame = pd.read_csv(filename)
frame = featurize(frame)
frame = filter_4(frame)
y = 'log(wage)'
x = ['education', 'experience', 'experience^2/100', 'intercept']
reg = regress.Reg(x, y)
reg.fit(frame)
reg.summarize() # OUT: equation 7.31, page 236
print('\nvce(hc2): ')
vce = pd.DataFrame(reg.vce('hc2'), index=reg.x_cols, columns=reg.x_cols)
print(vce) # OUT: equation 7.32, page 236
# Check that nlcom returns reasonable results
gradient = lambda beta: [100, 0, 0, 0]
print('\ns(theta_1):', nlcom.std_err(reg, gradient=gradient)) # OUT: ~0.8
gradient = lambda beta: [0, 100, 20, 0]
print('s(theta_2):', nlcom.std_err(reg, gradient=gradient))
# NOTE: the preceding result does not agree with Hansen's answer
gradient = lambda beta: [
0, (-50 / beta[2]), (50 * beta[1] / (beta[2]**2)), 0
]
print('s(theta_3):', nlcom.std_err(reg, gradient=gradient)) # OUT: ~7.0
# Check that htest.Wald returns reasonable results
R = np.array([[100, 0, 0, 0], [0, 100, 20, 0]])
theta = np.matmul(R, reg.beta)
gradient = lambda beta: np.transpose(R)
vce = nlcom.vce(reg, gradient=gradient) # [[0.632, 0.103], [0.103, 0.157]]
stat = htest.Wald(theta, vce=vce)
print('Wald (stat):', stat.value)
print('Wald (p-value):', stat.p_value())