def test_capm_lower_rf(series):
manager, index, rf = statistics.extract_returns_bmark_rf_partial(
series, 0, 7, 9, True, 0, True)
alpha, beta, r2 = statistics.capm(manager, index, rf)
assert alpha == pytest.approx(-0.0119, abs=MINE)
assert beta == pytest.approx(0.3991, abs=MINE)
assert r2 == pytest.approx(0.6079, abs=MINE)
def test_capm_nas(series):
manager, index, rf = statistics.extract_returns_bmark_rf(
series, 1, 7, None, True)
alpha, beta, r2 = statistics.capm(manager, index)
assert alpha == pytest.approx(0.0111, abs=MINE)
assert beta == pytest.approx(0.3431, abs=MINE)
assert r2 == pytest.approx(0.1704, abs=MINE)
def test_capm_lower(series):
manager, index, rf = statistics.extract_returns_bmark_rf_partial(
series, 0, 7, None, True, 0, True)
alpha, beta, r2 = statistics.capm(manager, index)
assert alpha == pytest.approx(-0.0100, abs=MINE)
assert beta == pytest.approx(0.4000, abs=MINE)
assert r2 == pytest.approx(0.6103, abs=MINE)
def test_capm_upper_rf(series):
manager, index, rf = statistics.extract_returns_bmark_rf_partial(
series, 0, 7, 9, True, 0, False)
alpha, beta, r2 = statistics.capm(manager, index, rf)
assert alpha == pytest.approx(0.0151, abs=MINE)
assert beta == pytest.approx(0.1992, abs=MINE)
assert r2 == pytest.approx(0.1985, abs=MINE)
def test_capm_rf(series):
manager, index, rf = statistics.extract_returns_bmark_rf(
series, 0, 7, 9, True)
alpha, beta, r2 = statistics.capm(manager, index, rf)
assert alpha == pytest.approx(0.0057, abs=MINE)
assert beta == pytest.approx(0.3900, abs=MINE)
assert r2 == pytest.approx(0.4338, abs=MINE)
def capm_table(data_series, manager_cols, index_col, rf_col):
# check to make sure nothing is missing
if index_col is None:
raise ValueError("Index Column cannot be blank")
if isinstance(index_col, collections.Iterable):
raise ValueError("Index Column must be a single value")
if manager_cols is None:
raise ValueError("Manager column cannot be None")
if not isinstance(manager_cols, collections.Iterable):
manager_cols = [manager_cols]
if rf_col is None:
raise ValueError("Risk free column cannot be None")
if isinstance(rf_col, collections.Iterable):
raise ValueError("Risk free column must be a single value")
# now create lists of the data points by comparing the manager to each comparison
# TODO this is really ugly, fix
alpha = []
beta = []
betap = []
betam = []
r2 = []
r = []
rp = []
te = []
ap = []
ir = []
tr = []
for mc in manager_cols:
manager, index, rf = pas.extract_returns_bmark_rf(data_series, mc, index_col, rf_col)
manager_u, index_u, rf_u = pas.extract_returns_bmark_rf_partial(data_series, mc, index_col, rf_col, lower=False)
manager_d, index_d, rf_d = pas.extract_returns_bmark_rf_partial(data_series, mc, index_col, rf_col)
alpha.append(pas.capm(manager, index, rf)[0])
beta.append(pas.capm(manager, index, rf)[1])
r2.append(pas.capm(manager, index, rf)[2])
betap.append(pas.capm(manager_u, index_u, rf_u)[1])
betam.append(pas.capm(manager_d, index_d, rf_d)[1])
r.append(pas.correlation(manager, index)[0])
rp.append(pas.correlation(manager, index)[1])
te.append(pas.tracking_error(manager, index))
ap.append(pas.active_premium(manager, index))
ir.append(pas.information_ratio(manager, index))
tr.append(pas.treynor_ratio(manager, index, rf))
st_data = {'Alpha': alpha,
'Beta': beta,
'Beta+': betap,
'Beta-': betam,
'R2': r2,
'Correlation': r,
'Correlation p-value': rp,
'Tracking Error': te,
'Active Premium': ap,
'Information Ratio': ir,
'Treynor Ratio': tr}
df = pd.DataFrame.from_dict(st_data, orient='index')
# replace the column names
colnames = [' vs '.join([data_series.columns[x], data_series.columns[index_col]]) for x in manager_cols]
# rename the columns
df = replace_col_names(df, colnames)
return df
data.even.apply(position)).dropna()
# strategy 2: short on even days, long on odd days
strategy2_return = (data['Close'].pct_change() *
-data.even.apply(position)).dropna()
strategy1_return.describe()
strategy2_return.describe()
pas.vol(snp500_return)
pas.beta(strategy1_return, snp500_return)
pas.beta(strategy2_return, snp500_return)
pas.correlation(strategy1_return, snp500_return)
pas.sharpe_ratio(snp500_return, 0.03)
pas.sharpe_ratio(strategy1_return, 0.03)
pas.sharpe_ratio(strategy2_return, 0.03)
pas.capm(strategy1_return, snp500_return)
pas.capm(strategy2_return, snp500_return)
pas.annualized_return(snp500_return, datetime.date(1928, 1, 3),
datetime.date(2020, 2, 12))
pas.annualized_return(strategy1_return, datetime.date(1928, 1, 3),
datetime.date(2020, 2, 12))
pas.annualized_return(strategy2_return, datetime.date(1928, 1, 3),
datetime.date(2020, 2, 12))
data['snp500_return'] = data['Close'].pct_change()
pat.stats_table(data, manager_col=8)
data['strategy1_return'] = (data['Close'].pct_change() *
data.even.apply(position))
pat.stats_table(data, manager_col=9)