def ninth_sector_comps():
""" Create list of companies composing the 9th sector """
# Load dataframe, 8 sector and 9 sector fits
X = pd.read_pickle('data/data.pkl')
XC8,S8,C8,SSE8,varexpl8 = local_load('saves/8_factor_matrices.pkl.gz')
XC9,S9,C9,SSE9,varexpl9 = local_load('saves/9_factor_matrices.pkl.gz')
# Divide companies by strongest sector affiliation for the 8 factor model
divided_comps_8 = []
for i in xrange(8):
index = np.where(np.argmax(S8,axis=0)==i)
comps = X.columns.values[index]
divided_comps_8.append(comps)
# Divide companies by strongest sector affiliation for the 9 factor model
divided_comps_9 = []
for i in xrange(9):
index = np.where(np.argmax(S9,axis=0)==i)
comps = X.columns.values[index]
divided_comps_9.append(comps)
# Determine overlap between 8 and 9 sector decompositions
count = np.zeros((9,8))
for i in xrange(9):
for j in xrange(8):
C = Counter(divided_comps_8[j]) & Counter(divided_comps_9[i])
count[i,j]=(len(list(C.elements())))
# Visually identified 9th sector and it's constituent companies from count
ninth_sector_companies = divided_comps_9[4]
return ninth_sector_companies
def top_three_sector_comps():
"""
Creates list of top 20 companies composing each of the sectors in the
three sector decomposition with the percentage they contribute to the sector
"""
X = pd.read_pickle('data/data.pkl')
# Add the sector and ticker information
sects = ['basic', 'capital', 'cyclical', 'energy', 'fin', 'health', 'noncyc',
'tech', 'telecom', 'utils', 'miscservices', 'realestate', 'retail', 'transport']
numcos = [58, 61, 41, 42, 107, 53, 40, 93, 6, 57, 55, 31, 46, 15]
sulis_sects = [[v]*k for k, v in zip(numcos, sects)]
sectsarray= np.array([item for sublist in sulis_sects for item in sublist])
tickers = X.columns.values.tolist()
sectors = sectsarray.tolist()
tuples = list(zip(*[tickers,sectors]))
index = pd.MultiIndex.from_tuples(tuples, names=['tickers', 'sectors'])
Xst = pd.read_pickle('data/data.pkl')
Xst.columns=index
# Load three factor matrices
XC3,S3,C3,SSE3,varexpl3 = local_load('saves/3_factor_matrices.pkl.gz')
# Determine top 20 companies in each of the three sectors
# and the percentage of the sector attributable to them
Corner1 = Xst.columns[C3[:,0].argsort()[-20:][::-1]]
Percent1 = np.round(C3[C3[:,0].argsort()[-20:][::-1]][:,0]*100,2)
Corner2 = Xst.columns[C3[:,1].argsort()[-20:][::-1]]
Percent2 = np.round(C3[C3[:,1].argsort()[-20:][::-1]][:,1]*100,2)
Corner3 = Xst.columns[C3[:,2].argsort()[-20:][::-1]]
Percent3 = np.round(C3[C3[:,2].argsort()[-20:][::-1]][:,2]*100,2)
return Corner1, Percent1, Corner2, Percent2, Corner3, Percent3
def company_pies():
""" Plot companies as pies colored by their exposure to each of the eight sectors"""
# Define colors, define labels and load data
cs = np.asarray([
'#984EA3', '#E41A1C', '#A65628', '#F781BF', '#4DAF4A', '#377EB8',
'#FF7F00', '#FFF000'
])
companies = np.asarray([
'Berry Petroleum', 'Corning', 'IBM', 'PG&E', 'Plum Creek Timber',
'Wal-Mart'
]).reshape((2, 3))
company_tickers = np.asarray(['BRY', 'GLW', 'IBM', 'PCG', 'PCL',
'WMT']).reshape((2, 3))
E, W, C, SSE, varexpl = local_load('saves/8_factor_matrices.pkl.gz')
X = pd.read_pickle('data/data.pkl')
# Plot
plt.figure(figsize=(9, 6))
plt.rcParams['patch.linewidth'] = 0
the_grid = gridspec.GridSpec(2, 3)
for i in xrange(2):
for j in xrange(3):
plt.subplot(the_grid[i, j], aspect=1)
index = np.argwhere(X.columns.values == company_tickers[i,
j])[0][0]
plt.pie(W.T[index], colors=cs, shadow=True, explode=[0.1] * 8)
plt.title(companies[i, j], fontsize=16)
plt.savefig('figures/company_pies.png', bbox_inches='tight', pad_inches=0)
plt.close()
return
def unweighted_price_index():
""" Create figure showing the unweighted price index of each sector """
E, W, C, SSE, varexpl = local_load('saves/8_factor_matrices.pkl.gz')
X_raw = np.load('data/raw_data.npy')
indxs = np.dot(X_raw, C)
indxs = indxs / indxs[0, :]
cs = np.asarray([
'#984EA3', '#E41A1C', '#A65628', '#F781BF', '#4DAF4A', '#377EB8',
'#FF7F00', '#FFF000'
])
order = [3, 1, 7, 0, 4, 2, 5, 6]
indxs = indxs[:, order]
cs = cs[order]
fig, axs = plt.subplots(1, 8, figsize=(12, 2.5))
for i in xrange(8):
plt.rcParams['ytick.color'] = '#636363'
plt.rcParams['xtick.color'] = '#636363'
plt.rcParams['axes.linewidth'] = 0
output = StringIO()
plt.ioff()
frameon = False
ax = plt.subplot(1, 8, i + 1)
plt.plot(indxs[:, i], color=cs[i])
ax.tick_params(axis='y', direction='out')
ax.yaxis.tick_left()
ax.tick_params(axis='x', direction='out')
ax.xaxis.tick_bottom()
plt.ylim(0, 10)
plt.xlim(0, 5001)
if i != 0:
plt.yticks([0, 5, 10], [])
else:
plt.yticks([0, 5, 10], [0, 5, 10])
if i < 0:
plt.xticks([343, 1853, 3359, 4869], [])
else:
plt.xticks([343, 1853, 3359, 4869], ['95', '01', '07', '13'])
plt.subplots_adjust(hspace=0.15, wspace=0.35)
plt.savefig('figures/unweighted_price_index.png',
bbox_inches='tight',
pad_inches=0)
plt.close()
return
def cumulative_log_returns():
""" Create figure showing the cumulative log returns of each sector """
E, W, C, SSE, varexpl = local_load('saves/8_factor_matrices.pkl.gz')
e = np.exp(E)
x = np.zeros([5001, 8])
x[1:, :] = E
x[0, :] = [1.] * 8
p = np.cumsum(x, axis=0)
indxs = (p / (250.0**.5))
order = [3, 1, 7, 0, 4, 2, 5, 6]
indxs = indxs[:, order]
cs = np.asarray([
'#984EA3', '#E41A1C', '#A65628', '#F781BF', '#4DAF4A', '#377EB8',
'#FF7F00', '#FFF000'
])
cs = cs[order]
fig, axs = plt.subplots(2, 4, figsize=(9.6, 6))
for i in xrange(8):
plt.rcParams['ytick.color'] = '#636363'
plt.rcParams['xtick.color'] = '#636363'
plt.rcParams['axes.linewidth'] = 0
output = StringIO()
plt.ioff()
frameon = False
ax = plt.subplot(2, 4, i + 1)
plt.plot(indxs[:, i], color=cs[i])
ax.tick_params(axis='y', direction='out')
ax.yaxis.tick_left()
ax.tick_params(axis='x', direction='out')
ax.xaxis.tick_bottom()
plt.xlim(0, 5001)
if i != 0 and i != 4:
plt.yticks([-2, 0, 2, 4], [])
else:
plt.yticks([-2, 0, 2, 4], [-2, 0, 2, 4])
if i < 4:
plt.xticks([343, 1853, 3359, 4869], [])
else:
plt.xticks([343, 1853, 3359, 4869], ['95', '01', '07', '13'])
plt.subplots_adjust(hspace=0.15, wspace=0.35)
plt.savefig('figures/cumulative_log_returns.png',
bbox_inches='tight',
pad_inches=0)
plt.close()
return
def tetrahedron_array_st():
"""
Make figure of tetrahedra along
different combinations of singular vectors
colored by scottrade classification
"""
# Load Fits
filename = 'saves/8_factor_matrices.pkl.gz'
E, W, C, SSE, varexpl = local_load(filename)
# Calculate Singular Vectors
A = np.dot(E, W)
u, s, vt = svd(A.T, full_matrices=0)
vtE = np.asarray(np.dot(vt, E))
# Define colors and labels
cols = [
'#FFF000', '#FDBF6F', '#A6CEE3', '#33A02C', '#FB9A99', '#B2DF8A',
'#E31A1C', '#6A3D9A', '#FF7F00', '#543005', '#CAB2D6', '#000000',
'#1F78B4', '#8C510A'
]
numcos = [58, 61, 41, 42, 107, 53, 40, 93, 6, 57, 55, 31, 46, 15]
sulis = [[v] * k for k, v in zip(numcos, cols)]
csuarray = np.array([item for sublist in sulis for item in sublist])
# Plot figure
fig, axs = plt.subplots(8, 8, figsize=(20, 20))
for ((j, i), ax) in np.ndenumerate(axs):
if i < j:
xs, ys = u[:, i].T * s[i], u[:, j].T * s[j]
ax.scatter(xs, ys, s=5, color=csuarray)
ax.scatter(vtE[i, :], vtE[j, :], s=10, color='k')
ax.set_axis_off()
xy = zip(vtE[i, :], vtE[j, :])
for p in itertools.combinations(xy, 2):
ax.plot([p[0][0], p[1][0]], [p[0][1], p[1][1]],
ls='--',
color='#D9D9D9')
else:
plt.delaxes(ax)
plt.subplots_adjust(hspace=0.05, wspace=0.05)
plt.savefig('figures/tetrahedron_array_st.png',
bbox_inches='tight',
pad_inches=0)
plt.close()
return
def single_tetrahedron():
"""
Make figure of single tetrahedron along 1st two singular vectors
colored by sector association
"""
# Load Fits
filename = 'saves/8_factor_matrices.pkl.gz'
E, W, C, SSE, varexpl = local_load(filename)
# Calculate Singular Vectors
A = np.dot(E, W)
u, s, vt = svd(A.T, full_matrices=0)
vtE = np.asarray(np.dot(vt, E))
# Define colors and assign to data points
csu = np.array([
'#984EA3', '#E41A1C', '#A65628', '#F781BF', '#4DAF4A', '#377EB8',
'#FF7F00', '#FFF000'
])
csuarray = csu[np.argmax(W, 0)]
# Initialize Plot
i = 0
j = 1
matplotlib.rcParams['ytick.color'] = '#636363'
matplotlib.rcParams['xtick.color'] = '#636363'
matplotlib.rcParams['axes.linewidth'] = 0
output = StringIO()
plt.ioff()
frameon = False
# Plot stock data and corner points
xs, ys = u[:, i].T * s[i], u[:, j].T * s[j]
plt.scatter(xs, ys, s=5, color=csuarray)
plt.scatter(vtE[i, :], vtE[j, :], s=10, color='k')
# Plot tetrathedron edges
xy = zip(vtE[i, :], vtE[j, :])
for p in itertools.combinations(xy, 2):
plt.plot([p[0][0], p[1][0]], [p[0][1], p[1][1]],
ls='-',
color='#D9D9D9')
# Save figure
plt.savefig('figures/single_tetrahedron.png',
bbox_inches='tight',
pad_inches=0)
plt.close()
return
def tetrahedron_array():
"""
Make figure of tetrahedra along
different combinations of singular vectors
colored by sector association
"""
# Load Fits
filename = 'saves/8_factor_matrices.pkl.gz'
E, W, C, SSE, varexpl = local_load(filename)
# Calculate Singular Vectors
A = np.dot(E, W)
u, s, vt = svd(A.T, full_matrices=0)
vtE = np.asarray(np.dot(vt, E))
# Define colors and assign to data points
csu = np.array([
'#984EA3', '#E41A1C', '#A65628', '#F781BF', '#4DAF4A', '#377EB8',
'#FF7F00', '#FFF000'
])
csuarray = csu[np.argmax(W, 0)]
# Plot
fig, axs = plt.subplots(8, 8, figsize=(20, 20))
for ((j, i), ax) in np.ndenumerate(axs):
if i < j:
xs, ys = u[:, i].T * s[i], u[:, j].T * s[j]
ax.scatter(xs, ys, s=5, color=csuarray)
ax.scatter(vtE[i, :], vtE[j, :], s=10, color='k')
ax.set_axis_off()
xy = zip(vtE[i, :], vtE[j, :])
for p in itertools.combinations(xy, 2):
ax.plot([p[0][0], p[1][0]], [p[0][1], p[1][1]],
ls='--',
color='#D9D9D9')
else:
plt.delaxes(ax)
plt.subplots_adjust(hspace=0.05, wspace=0.05)
plt.savefig('figures/tetrahedron_array.png',
bbox_inches='tight',
pad_inches=0)
plt.close()
return
def W():
"""
Weight distribution in canonical sectors. Each of the eight subplots shows the
constituent participation weights of all 705 companies in a canonical sector (rows of W_fs).
Stocks are colored by their scottrade classification.
"""
E, W, C, SSE, varexpl = local_load('saves/8_factor_matrices.pkl.gz')
cols = [
'#984ea3', '#6a51a3', '#df65b0', '#e41a1c', '#fff000', '#b3de69',
'#4daf4a', '#377eb8', '#006d2c', '#ff7f00', '#ae017e', '#a65628',
'#f781bf', '#3f007d'
]
sects = [
'basic', 'capital', 'cyclical', 'energy', 'fin', 'health', 'noncyc',
'tech', 'telecom', 'utils', 'miscservices', 'realestate', 'retail',
'transport'
]
numcos = [58, 61, 41, 42, 107, 53, 40, 93, 6, 57, 55, 31, 46, 15]
plt.rcParams['ytick.color'] = '#636363'
plt.rcParams['xtick.color'] = '#636363'
plt.rcParams['axes.linewidth'] = 0
order = [3, 1, 7, 0, 4, 2, 5, 6]
fig = plt.figure(figsize=(12, 12))
fig.frameon = False
for i in xrange(0, np.shape(W)[0], 1):
plt.subplot(5, 2, i + 1)
plt.xlim(0, 705)
plt.yticks([])
plt.ylim(0, 1)
if i > 5:
plt.xticks(xrange(0, 705, 100), [])
else:
plt.xticks(xrange(0, 705, 100), [])
plot_eve(W.T, order[i], numcos, sects, cols)
plt.subplots_adjust(hspace=0.2, wspace=0.1)
plt.savefig('figures/W.png', bbox_inches='tight', pad_inches=0)
plt.close()
return
def C():
"""
Creates Figure B1 in associated paper. See paper for description.
"""
E, W, C, SSE, varexpl = local_load('saves/8_factor_matrices.pkl.gz')
cols = [
'#984ea3', '#6a51a3', '#df65b0', '#e41a1c', '#fff000', '#b3de69',
'#4daf4a', '#377eb8', '#006d2c', '#ff7f00', '#ae017e', '#a65628',
'#f781bf', '#3f007d'
]
sects = [
'basic', 'capital', 'cyclical', 'energy', 'fin', 'health', 'noncyc',
'tech', 'telecom', 'utils', 'miscservices', 'realestate', 'retail',
'transport'
]
numcos = [58, 61, 41, 42, 107, 53, 40, 93, 6, 57, 55, 31, 46, 15]
plt.rcParams['ytick.color'] = '#636363'
plt.rcParams['xtick.color'] = '#636363'
plt.rcParams['axes.linewidth'] = 0
order = [3, 1, 7, 0, 4, 2, 5, 6]
fig = plt.figure(figsize=(12, 12))
fig.frameon = False
for i in xrange(0, np.shape(W)[0], 1):
plt.subplot(5, 2, i + 1)
plt.xlim(0, 705)
if i > 5:
plt.xticks(xrange(0, 701, 100),
['0', '100', '200', '300', '400', '500', '600', '700'])
else:
plt.xticks(xrange(0, 701, 100), [])
plot_eve(C, order[i], numcos, sects, cols)
plt.subplots_adjust(hspace=0.2, wspace=0.1)
plt.savefig('figures/C.png', bbox_inches='tight', pad_inches=0)
plt.close()
return
def noisy_flows():
""" Plots 'evolution' of noise applied to a static company decomposition """
flows = local_load("saves/noisy_flows.pkl.gz")
dataframe = pd.read_pickle('data/data.pkl')
companies = [
'Berry Petroleum', 'Corning', 'IBM', 'PG&E', 'Plum Creek Timber',
'Wal-Mart'
]
company_tickers = ['BRY', 'GLW', 'IBM', 'PCG', 'PCL', 'WMT']
for i in xrange(len(companies)):
ticks = dataframe.columns
tick = company_tickers[i]
plot_flows(flows, ticks, tick, plot_ts=0)
plt.savefig('figures/' + companies[i] + '_noisy_flow.png',
bbox_inches='tight',
pad_inches=0)
plt.close()
return
def flows():
""" Plots evolution of company sector associated in time """
flows = local_load("saves/flows.pkl.gz")
dataframe = pd.read_pickle('data/data.pkl')
companies = [
'Berry Petroleum', 'Corning', 'IBM', 'PG&E', 'Plum Creek Timber',
'Wal-Mart'
]
company_tickers = ['BRY', 'GLW', 'IBM', 'PCG', 'PCL', 'WMT']
for i in xrange(len(companies)):
ticks = dataframe.columns
tick = company_tickers[i]
plot_flows(flows, ticks, tick, plot_ts=0)
plt.savefig('figures/' + companies[i] + '_flow.png',
bbox_inches='tight',
pad_inches=0)
plt.close()
return
def fama_french():
"""
Produces projection plots for our three dimensional decomposition
and that of Fama and French colored by market cap values
"""
dataframe = pd.read_pickle('data/data.pkl') # log returns
data, MM = center_normalize_removemm(dataframe)
S3 = local_load('saves/3_factor_matrices_mm_removed.pkl.gz')[1]
#Load the fama and french data into pandas with the right parsers
ff_df = pd.io.parsers.read_table('data/FF_Factors.txt',
skiprows=4,
delimiter='\s+',
parse_dates=True)
ff_df = ff_df / 100
#restrict fama and french to the dates of interest
mytimes = dataframe.index
small_ff_df = ff_df.loc[mytimes]
#Loads the data for each of the stocks above in a list
with gzip.open('data/market_cap_list.pkl.gz', 'rb') as f:
market_cap_list = pickle.load(f)
##clean up the data
stock_names = []
market_caps = []
mc_stocks = []
for i, cap in enumerate(market_cap_list):
if cap != 'N/A':
if "B" in cap:
market_caps.append(float(cap.replace("B", "")) * 10**9)
if "M" in cap:
market_caps.append(float(cap.replace("M", "")) * 10**6)
stock_names.append(dataframe.columns[i])
mc_stocks.append(S3.T[i])
stock_names = np.asarray(stock_names)
market_caps = np.asarray(market_caps)
mc_stocks = np.asarray(mc_stocks)
##divide into large cap and small cap
big_mc_stocks = []
big_stock_names = []
med_mc_stocks = []
med_stock_names = []
sm_mc_stocks = []
sm_stock_names = []
bigmed_cutoff = 10 * 10**9
smmed_cutoff = 2 * 10**9
for i in xrange(len(market_caps)):
if market_caps[i] >= bigmed_cutoff:
big_mc_stocks.append(mc_stocks[i])
big_stock_names.append(stock_names[i])
elif market_caps[i] <= smmed_cutoff:
sm_mc_stocks.append(mc_stocks[i])
sm_stock_names.append(stock_names[i])
else:
med_mc_stocks.append(mc_stocks[i])
med_stock_names.append(stock_names[i])
big_mc_stocks = np.asarray(big_mc_stocks)
med_mc_stocks = np.asarray(med_mc_stocks)
sm_mc_stocks = np.asarray(sm_mc_stocks)
size = 3
fig, axs = plt.subplots(size,
size,
figsize=(20, 20),
subplot_kw={"frameon": False},
sharex=True,
sharey=True)
for i in range(size):
for j in range(size):
if j > i:
k = list(
set([0, 1,
2]).difference(set([i])).difference(set([j])))[0]
c = np.r_[["#e41a1c"] * len(big_mc_stocks[:, i]),
["#377eb8"] * len(med_mc_stocks[:, i]),
["#4daf4a"] * len(sm_mc_stocks[:, i])]
xs = np.r_[big_mc_stocks[:, i], med_mc_stocks[:, i],
sm_mc_stocks[:, i]]
ys = np.r_[big_mc_stocks[:, j], med_mc_stocks[:, j],
sm_mc_stocks[:, j]]
zs = np.r_[big_mc_stocks[:, k], med_mc_stocks[:, k],
sm_mc_stocks[:, k]]
pks = zs.argsort()
axs[i, j].scatter(xs[pks],
ys[pks],
c=c[pks],
linewidth=0,
alpha=0.8)
axs[i, j].axis("equal")
else:
axs[i, j].axis('off')
plt.savefig('figures/fama_french1.png', bbox_inches='tight', pad_inches=0)
big_stocks = []
for name in big_stock_names:
big_stocks.append(dataframe[name])
med_stocks = []
for name in med_stock_names:
med_stocks.append(dataframe[name])
sm_stocks = []
for name in sm_stock_names:
sm_stocks.append(dataframe[name])
big_stocks = pd.DataFrame(big_stocks)
med_stocks = pd.DataFrame(med_stocks)
sm_stocks = pd.DataFrame(sm_stocks)
out = dolinearregression(big_stocks.T,
small_ff_df[['FFMkt-RF', 'FFSMB', 'FFHML']],
small_ff_df,
MM,
sub=True)
big_w = out[0]
out = dolinearregression(med_stocks.T,
small_ff_df[['FFMkt-RF', 'FFSMB', 'FFHML']],
small_ff_df,
MM,
sub=True)
med_w = out[0]
out = dolinearregression(sm_stocks.T,
small_ff_df[['FFMkt-RF', 'FFSMB', 'FFHML']],
small_ff_df,
MM,
sub=True)
sm_w = out[0]
size = 3
fig, axs = plt.subplots(size,
size,
figsize=(20, 20),
subplot_kw={"frameon": False},
sharex=True,
sharey=True)
for i in range(size):
for j in range(size):
if j > i:
#k = # other index
#sortpk = big_w[k,:].argsort()
k = list(
set([0, 1,
2]).difference(set([i])).difference(set([j])))[0]
c = np.r_[["#e41a1c"] * len(big_w[i, :]),
["#377eb8"] * len(med_w[i, :]),
["#4daf4a"] * len(sm_w[i, :])]
xs = np.r_[big_w[i, :], med_w[i, :], sm_w[i, :]]
ys = np.r_[big_w[j, :], med_w[j, :], sm_w[j, :]]
zs = np.r_[big_w[k, :], med_w[k, :], sm_w[k, :]]
pks = zs.argsort()
axs[i, j].scatter(xs[pks],
ys[pks],
c=c[pks],
linewidth=0,
alpha=0.8)
axs[i, j].axis("equal")
else:
axs[i, j].axis('off')
plt.savefig('figures/fama_french2.png', bbox_inches='tight', pad_inches=0)
plt.close()
return
def sector_pies():
"""
Plot sectors as pies colored by their
relation to a sector decomposition with
a different number of archetypes
"""
# Load in fit data
E_list = []
weights = []
for i in xrange(2, 10):
E, W, C, SSE, varexpl = local_load('saves/' + str(i) +
'_factor_matrices.pkl.gz')
E_list.append(E)
if i < 9:
weights.append(W)
if i == 8:
C8 = C
E2 = E_list[0]
E3 = E_list[1]
E8 = E_list[-2]
# Determine relationship between decompositions with varying numbers of sectors
E, W, SSE, varexpl = archanalysis_fixed_C(E2, E8, noc=8, i_0=0)
W28 = W
E, W, SSE, varexpl = archanalysis_fixed_C(E3, E8, noc=8, i_0=0)
W38 = W
E, W, SSE, varexpl = archanalysis_fixed_C(E8, E2, noc=2, i_0=0)
W82 = W
E, W, SSE, varexpl = archanalysis_fixed_C(E8, E3, noc=3, i_0=0)
W83 = W
# Determine colors
w_list = []
for j in xrange(len(E_list) - 2):
XC, S, SSE, varexpl = archanalysis_fixed_C(E_list[j],
E_list[j + 1],
noc=j + 3,
i_0=0)
W = S
w = (W / W.sum(axis=0)[None, :]).T #normalized
w_list.append(w)
num_comps = []
for i in xrange(len(weights)):
temp_arry = []
for j in xrange(i + 2):
temp_arry.append(
len(np.where(np.argmax(weights[i], axis=0) == j)[0]))
num_comps.append(np.asarray(temp_arry))
cs_list = []
cs = np.asarray([
'#984EA3', '#E41A1C', '#A65628', '#F781BF', '#4DAF4A', '#377EB8',
'#FF7F00', '#FFF000'
])
cs_list.append(cs)
for i in xrange(1, 7):
cs = define_colors(cs, w_list[-i], num_comps[-i])
cs_list.append(cs)
cs_list = cs_list[::-1]
# Plot pies
noc = 2
plt.figure(figsize=(6, 3))
the_grid = gridspec.GridSpec(1, noc)
sectors2 = ['c-assets', 'c-goods']
for j in xrange(noc):
plt.subplot(the_grid[0, j], aspect=1)
plt.pie(W28.T[j], colors=cs_list[-1], shadow=True)
plt.title(sectors2[j], fontsize=20)
plt.savefig('figures/W28.png', bbox_inches='tight', pad_inches=0)
plt.close()
noc = 3
plt.figure(figsize=(9, 3))
the_grid = gridspec.GridSpec(1, noc)
sectors3 = ['c-staples', 'c-assets', 'c-tech']
for j in xrange(noc):
plt.subplot(the_grid[0, j], aspect=1)
plt.pie(W38.T[j], colors=cs_list[-1], shadow=True)
plt.title(sectors3[j], fontsize=20)
plt.savefig('figures/W38.png', bbox_inches='tight', pad_inches=0)
plt.close()
sectors = np.asarray([
'c-industrial', 'c-energy', 'c-real estate', 'c-cyclical',
'c-non-cyclical', 'c-tech', 'c-utility', 'c-financial'
])
noc = 8
plt.figure(figsize=(24, 3))
the_grid = gridspec.GridSpec(1, noc)
for j in xrange(noc):
plt.subplot(the_grid[0, j], aspect=1)
plt.pie(W82.T[j], colors=cs_list[0], shadow=True)
plt.title(sectors[j], fontsize=20)
plt.savefig('figures/W82.png', bbox_inches='tight', pad_inches=0)
plt.close()
noc = 8
plt.figure(figsize=(24, 3))
the_grid = gridspec.GridSpec(1, noc)
for j in xrange(noc):
plt.subplot(the_grid[0, j], aspect=1)
plt.pie(W83.T[j], colors=cs_list[1], shadow=True)
plt.title(sectors[j], fontsize=20)
plt.savefig('figures/W83.png', bbox_inches='tight', pad_inches=0)
plt.close()
return
def single_tetrahedron_st():
"""
Make figure of single tetrahedron along
1st two singular vectors
colored by scottrade classification
"""
# Load Fits
filename = 'saves/8_factor_matrices.pkl.gz'
E, W, C, SSE, varexpl = local_load(filename)
# Calculate Singular Vectors
A = np.dot(E, W)
u, s, vt = svd(A.T, full_matrices=0)
vtE = np.asarray(np.dot(vt, E))
#Define colors and labels
cols = [
'#FFF000', '#FDBF6F', '#A6CEE3', '#33A02C', '#FB9A99', '#B2DF8A',
'#E31A1C', '#6A3D9A', '#FF7F00', '#543005', '#CAB2D6', '#000000',
'#1F78B4', '#8C510A'
]
sects = [
'basic', 'capital', 'cyclical', 'energy', 'fin', 'health', 'noncyc',
'tech', 'telecom', 'utils', 'miscservices', 'realestate', 'retail',
'transport'
]
numcos = [58, 61, 41, 42, 107, 53, 40, 93, 6, 57, 55, 31, 46, 15]
sulis = [[v] * k for k, v in zip(numcos, cols)]
csuarray = np.array([item for sublist in sulis for item in sublist])
sulis_sects = [[v] * k for k, v in zip(numcos, sects)]
sectsarray = np.array(
[item for sublist in sulis_sects for item in sublist])
# Plot figure
matplotlib.rcParams['ytick.color'] = '#636363'
matplotlib.rcParams['xtick.color'] = '#636363'
matplotlib.rcParams['axes.linewidth'] = 0
output = StringIO()
plt.ioff()
frameon = False
i = 0
j = 1
xs, ys = u[:, i].T * s[i], u[:, j].T * s[j]
plt.scatter(xs, ys, s=5, color=csuarray)
plt.savefig('figures/single_tetrahedron_st.png',
bbox_inches='tight',
pad_inches=0)
plt.clf()
# Make a legend
recs = []
for i in range(0, len(cols)):
recs.append(mpatches.Rectangle([0, 0], 1, 1, fc=cols[i]))
plt.legend(recs, sects)
plt.axis('off')
plt.savefig('figures/single_tetrahedron_st_legend.png',
bbox_inches='tight',
pad_inches=0)
plt.close()
return
