def run_regressions(totals: pd.DataFrame,
window: int = 3,
infectious_period: float = 4.5) -> pd.DataFrame:
# run rolling regressions and get parameters
model = RollingOLS.from_formula(formula="logdelta ~ time",
window=window,
data=totals)
rolling = model.fit(method="lstsq")
growthrates = rolling.params.join(rolling.bse, rsuffix="_stderr")
growthrates["rsq"] = rolling.rsquared
growthrates.rename(
lambda s: s.replace("time", "gradient").replace("const", "intercept"),
axis=1,
inplace=True)
# calculate growth rates
growthrates[
"egrowthrateM"] = growthrates.gradient + 2 * growthrates.gradient_stderr
growthrates[
"egrowthratem"] = growthrates.gradient - 2 * growthrates.gradient_stderr
growthrates["R"] = growthrates.gradient * infectious_period + 1
growthrates[
"RM"] = growthrates.gradient + 2 * growthrates.gradient_stderr * infectious_period + 1
growthrates[
"Rm"] = growthrates.gradient - 2 * growthrates.gradient_stderr * infectious_period + 1
growthrates["date"] = growthrates.index
growthrates["days"] = totals.time
return growthrates
def test_formula():
y, x, w = gen_data(250, 3, True, pandas=True)
fmla = "y ~ 1 + x0 + x1 + x2"
data = pd.concat([y, x], axis=1)
mod = RollingWLS.from_formula(fmla, window=100, data=data, weights=w)
res = mod.fit()
alt = RollingWLS(y, x, window=100)
alt_res = alt.fit()
assert_allclose(res.params, alt_res.params)
ols_mod = RollingOLS.from_formula(fmla, window=100, data=data)
ols_mod.fit()
def rollingOLS(totals: pd.DataFrame, window: int = 3, infectious_period: float = 4.5) -> pd.DataFrame:
""" legacy rolling regression-based implementation of Bettencourt/Ribeiro method """
# run rolling regressions and get parameters
model = RollingOLS.from_formula(formula = "logdelta ~ time", window = window, data = totals)
rolling = model.fit(method = "lstsq")
growthrates = rolling.params.join(rolling.bse, rsuffix="_stderr")
growthrates["rsq"] = rolling.rsquared
growthrates.rename(lambda s: s.replace("time", "gradient").replace("const", "intercept"), axis = 1, inplace = True)
# calculate growth rates
growthrates["egrowthrateM"] = growthrates.gradient + 2 * growthrates.gradient_stderr
growthrates["egrowthratem"] = growthrates.gradient - 2 * growthrates.gradient_stderr
growthrates["R"] = growthrates.gradient * infectious_period + 1
growthrates["RM"] = growthrates.gradient + 2 * growthrates.gradient_stderr * infectious_period + 1
growthrates["Rm"] = growthrates.gradient - 2 * growthrates.gradient_stderr * infectious_period + 1
growthrates["date"] = growthrates.index.get_level_values('status_change_date')
growthrates["days"] = totals.time
return growthrates
result.dropna(
axis=0, how='any', inplace=True
) # not every month we have factor return "result.isnull().sum()"
final_result = pd.concat([final_result, longshort], axis=0)
# Step 4: Do risk adjustment in FF3F framework
# convert factor values to pure factor returns (without size or industry) by using long short just like ff3f
factor = pd.DataFrame({
'yearmonth': result['yearmonth'],
f: result['ret_f0f1_high'] - result['ret_f0f1_low']
})
df_return = pd.merge(df_return, factor, on='yearmonth')
model = RollingOLS.from_formula(f'{f}~mktrf+smb+hml',
data=df_return,
window=12).fit() # 1 year rolling
asset_pricing = pd.DataFrame([], index=[f + ' exposure', f + ' tstat'])
asset_pricing['alpha'] = [
np.mean(model.params['Intercept']),
np.mean(model.tvalues['Intercept'])
]
asset_pricing['beta_mktrf'] = [
np.mean(model.params['mktrf']),
np.mean(model.tvalues['mktrf'])
]
asset_pricing['beta_smb'] = [
np.mean(model.params['smb']),
np.mean(model.tvalues['smb'])
]
asset_pricing['beta_hml'] = [
exog = sm.add_constant(factors[exog_vars])
rols = RollingOLS(endog, exog, window=60)
rres = rols.fit()
fig = rres.plot_recursive_coefficient(variables=exog_vars, figsize=(14, 18))
# ## Formulas
#
# `RollingOLS` and `RollingWLS` both support model specification using the
# formula interface. The example below is equivalent to the 3-factor model
# estimated previously. Note that one variable is renamed to have a valid
# Python variable name.
joined = pd.concat([factors, industries], axis=1)
joined["Mkt_RF"] = joined["Mkt-RF"]
mod = RollingOLS.from_formula("HiTec ~ Mkt_RF + SMB + HML",
data=joined,
window=60)
rres = mod.fit()
rres.params.tail()
# ## `RollingWLS`: Rolling Weighted Least Squares
#
# The `rolling` module also provides `RollingWLS` which takes an optional
# `weights` input to perform rolling weighted least squares. It produces
# results that match `WLS` when applied to rolling windows of data.
# ## Fit Options
#
# Fit accepts other optional keywords to set the covariance estimator.
# Only two estimators are supported, `'nonrobust'` (the classic OLS
# estimator) and `'HC0'` which is White's heteroskedasticity robust
