def PlotConfidenceIntervals(xs,
inters,
slopes,
res=None,
percent=90,
**options):
"""Plots the 90% confidence intervals for weights based on ages.
xs: sequence
inters: estimated intercepts
slopes: estimated slopes
res: residuals
percent: what percentile range to show
"""
fys_seq = []
for inter, slope in zip(inters, slopes):
fxs, fys = thinkstats2.FitLine(xs, inter, slope)
if res is not None:
fys += np.random.permutation(res)
fys_seq.append(fys)
p = (100 - percent) / 2
percents = p, 100 - p
low, high = thinkstats2.PercentileRows(fys_seq, percents)
thinkplot.FillBetween(fxs, low, high, **options)
def ResampleSurvival(resp, iters=101):
"""Resamples respondents and estimates the survival function.
resp: DataFrame of respondents
iters: number of resamples
"""
_, sf = EstimateMarriageSurvival(resp)
thinkplot.Plot(sf)
low, high = resp.min(), resp.max()
ts = np.arange(low, high, 1/12.0)
ss_seq = []
for _ in range(iters):
sample = thinkstats2.ResampleRowsWeighted(resp)
_, sf = EstimateMarriageSurvival(sample)
ss_seq.append(sf.Probs(ts))
low, high = thinkstats2.PercentileRows(ss_seq, [5, 95])
thinkplot.FillBetween(ts, low, high, color='gray', label='90% CI')
thinkplot.Save(root='survival3',
xlabel='age (years)',
ylabel='prob unmarried',
xlim=[12, 46],
ylim=[0, 1],
formats=FORMATS)
def PlotSurvivalFunctions(sf_map, predict_flag=False):
"""Plot estimated survival functions.
sf_map: map from group name to sequence of survival functions
predict_flag: whether the lines are predicted or actual
"""
thinkplot.PrePlot(len(sf_map))
for name, sf_seq in sorted(sf_map.items(), reverse=True):
if len(sf_seq) == 0:
continue
sf = sf_seq[0]
if len(sf) == 0:
continue
ts, rows = MakeSurvivalCI(sf_seq, [10, 50, 90])
thinkplot.FillBetween(ts, rows[0], rows[2], color='gray')
if not predict_flag:
thinkplot.Plot(ts, rows[1], label='19%d'%name)
thinkplot.Config(xlabel='age (years)', ylabel='prob unmarried',
xlim=[14, 45], ylim=[0, 1],
legend=True, loc='upper right')
def SimulateAutocorrelation(daily, iters=1001, nlags=40):
"""Resample residuals, compute autocorrelation, and plot percentiles.
daily:
iters:
nlags:
"""
# run simulations
t = []
for i in range(iters):
filled = FillMissing(daily, span=30)
resid = thinkstats2.Resample(filled.resid)
acf = smtsa.acf(resid, nlags=nlags, unbiased=True)[1:]
t.append(np.abs(acf))
# put the results in an array and sort the columns
size = iters, len(acf)
array = np.zeros(size)
for i, acf in enumerate(t):
array[i, ] = acf
array = np.sort(array, axis=0)
# find the bounds that cover 95% of the distribution
high = PercentileRow(array, 97.5)
low = -high
lags = range(1, nlags + 1)
thinkplot.FillBetween(lags, low, high, alpha=0.2, color='gray')
def PlotSurvivalFunctions(sf_map, predict_flag=False, colormap=None):
"""Plot estimated survival functions.
sf_map: map from group name to sequence of survival functions
predict_flag: whether the lines are predicted or actual
colormap: map from group name to color
"""
thinkplot.PrePlot(num=len(sf_map))
for name, sf_seq in sorted(sf_map.items(), reverse=True):
if len(sf_seq) == 0:
continue
sf = sf_seq[0]
if len(sf) == 0:
continue
ts, rows = MakeSurvivalCI(sf_seq, [10, 50, 90])
thinkplot.FillBetween(ts, rows[0], rows[2], color='gray', alpha=0.2)
if not predict_flag:
if colormap:
color = colormap[name]
thinkplot.Plot(ts, rows[1], label='%ds' % name, color=color)
else:
thinkplot.Plot(ts, rows[1], label='%ds' % name)
def PlotConfidenceIntervals(xs, inters, slopes, percent=90, **options):
fys_seq = []
for inter, slope in zip(inters, slopes):
fxs, fys = thinkstats2.FitLine(xs, inter, slope)
fys_seq.append(fys)
p = (100 - percent) / 2
percents = p, 100 - p
low, high = thinkstats2.PercentileRows(fys_seq, percents)
thinkplot.FillBetween(fxs, low, high, **options)
def PlotPredictions(daily, years, iters=101, percent=90, func=RunLinearModel):
"""Plots predictions.
daily: DataFrame of daily prices
years: sequence of times (in years) to make predictions for
iters: number of simulations
percent: what percentile range to show
func: function that fits a model to the data
"""
result_seq = SimulateResults(daily, iters=iters, func=func)
p = (100 - percent) / 2
percents = p, 100 - p
predict_seq = GeneratePredictions(result_seq, years, add_resid=True)
low, high = thinkstats2.PercentileRows(predict_seq, percents)
thinkplot.FillBetween(years, low, high, alpha=0.3, color='gray')
predict_seq = GeneratePredictions(result_seq, years, add_resid=False)
low, high = thinkstats2.PercentileRows(predict_seq, percents)
thinkplot.FillBetween(years, low, high, alpha=0.5, color='gray')
def PlotIntervals(daily, years, iters=101, percent=90, func=RunLinearModel):
"""Plots predictions based on different intervals.
daily: DataFrame of daily prices
years: sequence of times (in years) to make predictions for
iters: number of simulations
percent: what percentile range to show
func: function that fits a model to the data
"""
result_seq = SimulateIntervals(daily, iters=iters, func=func)
p = (100 - percent) / 2
predictions = GeneratePredictions(result_seq, years, add_resid=True)
low = PercentileRow(predictions, p)
high = PercentileRow(predictions, 100 - p)
thinkplot.FillBetween(years, low, high, alpha=0.1, color='gray')
def SimulateAutocorrelation(daily, iters=1001, nlags=40):
"""Resample residuals, compute autocorrelation, and plot percentiles
daily: DataFrame
iters: number of simulations to run
nlags: maximum lags to compute autocorrelation
"""
t = []
for _ in range(iters):
filled = FillMissing(daily, span=30)
resid = thinkstats2.Resample(filled.resid)
acf = smtsa.acf(resid, nlags=nlags, unbiased=True)[1:]
t.append(np.abs(acf))
high = thinkstats2.PercentileRows(t, [97.5])[0]
low = -high
lags = range(1, nlags + 1)
thinkplot.FillBetween(lags, low, high, alpha=0.2, color='gray')
def ResampleSurvival(dados, limiar, iters=101):
"""Resamples respondents and estimates the survival function.
resp: DataFrame of respondents
iters: number of resamples
"""
_, sf = EstimateMarriageSurvival(dados, limiar)
thinkplot.Plot(sf)
low, high = dados.min(), dados.max()
ts = np.arange(low, high, 1)
ss_seq = []
for _ in range(iters):
sample = thinkstats2.ResampleRowsWeighted(pd.DataFrame(dados), column='MANSO')
_, sf = EstimateMarriageSurvival(sample['MANSO'], limiar)
ss_seq.append(sf.Probs(ts))
low, high = thinkstats2.PercentileRows(ss_seq, [5, 95])
thinkplot.FillBetween(ts, low, high, color='gray', label='90% CI')
def PlotConfidenceIntervals(xs,
inters,
slopes,
res=None,
percent=90,
**options):
"""Plots the 90% confidence intervals for weights based on ages.
xs: sequence
inters: estimated intercepts
slopes: estimated slopes
res: residuals
percent: what percentile range to show
"""
size = len(slopes), len(xs)
array = np.zeros(size)
for i, (inter, slope) in enumerate(zip(inters, slopes)):
fxs, fys = thinkstats2.FitLine(xs, inter, slope)
if res is not None:
fys += np.random.permutation(res)
array[i, ] = fys
array = np.sort(array, axis=0)
def Percentile(p):
"""Selects the line from array that corresponds to percentile p.
p: float 0--100
returns: NumPy array (one row)
"""
index = int(len(slopes) * p / 100)
return array[index, ]
p = (100 - percent) / 2
#low = thinkstats2.Smooth(Percentile(p))
#high = thinkstats2.Smooth(Percentile(100-p))
low = Percentile(p)
high = Percentile(100 - p)
thinkplot.FillBetween(fxs, low, high, **options)
