def tornadoPlot(self, simId, scenario, resultName, tornadoType,
sliderValue, selectedData):
corrDF = self.getCorrDF(simId, scenario, resultName)
if tornadoType == 'normalized':
squared = corrDF.spearman**2
corrDF['normalized'] = squared / squared.sum()
corrDF['sign'] = 1
corrDF.ix[(corrDF.spearman < 0), 'sign'] = -1
corrDF[
'value'] = corrDF.normalized * corrDF.sign # normalized squares with signs restored
plotColumn = 'value'
title = 'Normalized rank correlation'
else:
plotColumn = 'spearman'
title = 'Rank correlation'
title += ' of {} for scenario {}'.format(resultName, scenario)
varCount = min(
20, len(corrDF)
) # if sliderValue is None else len(corrDF[corrDF['abs'] >= sliderValue]))
varNames = list(reversed(corrDF.index[:varCount]))
values = list(reversed(corrDF[plotColumn][:varCount]))
plotData = go.Bar(
orientation='h',
x=values,
y=varNames,
width=0.6,
marker=dict(
color=getColor('TornadoPlotBar'),
line=dict(color=getColor('TornadoPlotLine'), width=0),
),
textposition='none',
name='Rank correlation',
hoverinfo='skip', # prevents hover events
)
layout = updateStyle('Plot',
title=title,
margin=updateStyle('PlotMargin', l=300),
xaxis=dict(range=[-1, 1]),
dragmode='select')
figure = dict(data=[plotData], layout=layout)
return figure
def showDistribution(project, simId, scenario, outputName, sliderInfo,
distOptions, selectedData):
_logger.debug('showDistribution(%s, %s, %s, %s, %s)' %
(project, simId, scenario, outputName, selectedData))
# Clear selectedData if context changes
context = (project, simId, scenario, outputName)
if not data.histogramContext or context != data.histogramContext:
_logger.debug('showDistribution: clearing context')
selectedData = sliderInfo = None
data.histogramContext = context
# if called before values are known, return an empty plot
if not (project and scenario and outputName) or simId is None:
plotData = {'x': []}
title = 'Please select a model output to plot'
annotations = None
else:
plotData, title, annotations = data.distPlot(
simId, scenario, outputName, sliderInfo, distOptions,
selectedData)
# TBD: generalize this
if Oct16:
if outputName == 'percent-change':
xtitle = 'Change from baseline fuel'
xticksuffix = '%'
else:
# latex formatting is broken, but HTML works fine
# xtitle = '$g CO_2 MJ^{-1}$' if Oct16 else ''
xtitle = 'g CO<sub>2</sub> MJ<sup>-1</sup>'
xticksuffix = None
else:
xtitle = data.db.getOutputUnits(outputName)
xticksuffix = ''
layout = updateStyle('Plot',
title=title,
yaxis={
'title': 'Probability density',
'tickvals': []
},
xaxis={
'title': xtitle,
'ticksuffix': xticksuffix,
},
margin=getStyle('PlotMarginWithXTitle'),
dragmode='select',
showLegend=True,
legend=dict(x=0.0,
y=1.0,
bgcolor=getColor('PlotBg')))
if annotations:
layout['annotations'] = annotations
figure = dict(data=plotData, layout=layout)
return figure
def layout(self):
layout = html.Div(
[
dataStore(id='paraCoords-vars'),
dataStore(id='dist-selected'),
html.H1('Monte Carlo Simulation Explorer', className='title'),
html.Table(
[
html.Tr([
html.Th('Project'),
html.Th('Simulation'),
html.Th('Scenario'),
html.Th('Model Output'),
]),
html.Tr([
html.Td(self.projectChooser()),
html.Td(self.simChooser()),
html.Td(self.scenarioChooser()),
html.Td(self.outputChooser())
])
],
style={
'width': '100%',
'border': '1px solid black',
'table-layout': 'fixed',
'background-color': getColor('ChooserBgColor')
}),
html.Div(
[
# distribution cluster and tornado chart
html.Div([
html.Div(self.distributionSectionLayout(),
className='cell twocol'),
html.Div(self.tornadoSectionLayout(),
className='cell twocol')
],
className='row'),
# parallel coordinates section
html.Div([
html.Div([
html.P([
'Brush vertically over ranges to filter; click outside the selected range to reset. ',
'Drag variable names below to reorder. Slider sets the number of vars to plot'
],
style=getStyle('HelpText')),
html.Div(
[
dcc.Slider(
id='paraCoords-slider',
min=1,
max=10,
step=1,
marks={
value:
sliderLabel(value)
for value in range(1, 11, 1)
},
updatemode='drag',
value=6),
],
style={
'margin-bottom': 30,
'margin-left': 'auto',
'margin-right': 'auto',
'width': 400
}),
html.Div(dcc.Graph(id='paraCoords')),
],
className='cell onecol')
],
className='row'),
# Correlation convergence section
html.Div([
html.Div([
html.Span('Correlation convergence',
style=getStyle('Label')),
html.Div('', style={'height': 15}),
html.Div(dcc.Graph(id='corr-convergence')),
],
className='cell onecol')
],
className='row'),
# Scatterplot section
html.Div([
html.Div([
html.Div([
html.Div(style={'height': 15}),
html.Span(
'Model Inputs (random variables)',
style=getStyle('Label')),
html.Div(style={'height': 15}),
self.inputChooser(multi=True)
],
style=updateStyle('Chooser',
width=900)),
html.Div(style={'height': 15}),
html.Div([
html.Span('Model Outputs',
style=getStyle('Label')),
self.multiOutputChooser()
],
style=updateStyle('Chooser',
width=900)),
html.Div([
html.Button('Create scatterplot',
id='scatterplot-button',
style=getStyle('Button'))
],
style={'margin': 5}),
html.Div([dcc.Graph(id='scatter-matrix')],
style=updateStyle('AutoMargins'))
],
className='cell onecol')
],
className='row'),
],
className='table',
style={
'width': 1200,
'margin-left': 'auto',
'margin-right': 'auto'
}),
],
style=getStyle('Page'))
return layout
def scatterPlots(self, simId, scenario, inputs, outputs):
"""
Plot a set of small scatterplots showing correlation between chosen
model inputs and outputs.
:param simId: (int) simulation id
:param scenario: (str) scenario name
:param inputs: (iterable of str) names of inputs to plot
:param outputs: (iterable of str) names of outputs to plot
:return: dash 'figure' data.
"""
allInputs = self.getParameterValues(simId)
inputsDF = allInputs[inputs]
outputsDF = pd.DataFrame(columns=outputs)
for output in outputs:
outputsDF[output] = self.getOutValues(simId, scenario, output)
inputsDF = inputsDF.iloc[
outputsDF.index] # select only trials for which we have results
numIns = len(inputs)
numOuts = len(outputs)
fig = tools.make_subplots(rows=numOuts,
cols=numIns,
vertical_spacing=0.01,
shared_xaxes=True,
shared_yaxes=True)
layout = fig['layout']
for row, output in enumerate(outputs):
yAxisNum = row + 1
xAxisNum = 0
for col, input in enumerate(inputs):
xAxisNum += 1
trace = go.Scatter(
x=inputsDF[input],
y=outputsDF[output],
hoverinfo='skip', # don't show or fire events
mode='markers',
marker=dict(size=1,
opacity=0.9,
color=getColor('ScatterPoints')))
fig.append_trace(trace, row + 1, col + 1)
# Compute the names of the corresponding axes
xaxis = "xaxis{}".format(xAxisNum)
yaxis = "yaxis{}".format(yAxisNum)
layout[xaxis].update(title=input,
showgrid=False,
zeroline=False)
layout[yaxis].update(title=output,
showgrid=False,
zeroline=False)
layout.update(
margin=dict(l=50, r=30, t=30, b=30),
height=140 * numOuts + 20,
#width= 130 * numIns,
showlegend=False,
font=dict(size=9, family=getFont('PlotText')))
return fig
def distPlot(self, simId, scenario, outputName, sliderInfo, distOptions,
selectedData):
"""
Generate the plot data for the distribution plot. Return three items:
the plot data, title, and annotations (might be empty list).
"""
annotations = []
colors = []
values = self.getOutValues(simId, scenario, outputName)
if values is None:
title = 'Please select a model output to plot'
plotData = [
dict(
type='bar',
x=[],
y=[],
name=outputName,
showlegend=False,
)
]
return plotData, title, annotations
title = 'Distribution of %s for scenario %s' % (outputName, scenario)
bins = min(
100,
len(values)) or 1 # 1 for corner case of no data; bins must be > 0
# Generate histogram data to be able to color bars directly
counts, edges = np.histogram(values, bins, density=True)
barCount = len(counts)
# Plot the counts on x-axis at the mean of each pair of edges
barValues = [np.mean(edges[i:i + 1]) for i in range(barCount)]
if sliderInfo:
minQ, maxQ = sliderInfo
minX = values.quantile(q=minQ / 100.0, interpolation='linear')
maxX = values.quantile(q=maxQ / 100.0, interpolation='linear')
else:
minX, maxX = selectedData['range']['x'] if selectedData else (
edges[0], edges[-1])
# Sub-select trials based on selection range to show in parallel coords plot
self.selectedResults = values[values.between(minX,
maxX,
inclusive=True)]
# highlight selected bars by desaturating non-selected items
active = getColor('DistBarsActive')
inactive = getColor('DistBarsInactive')
for i in range(barCount):
colors.append(active if minX <= barValues[i] <= maxX else inactive)
distPlot = ff.create_distplot([values], [outputName],
bin_size=1,
show_hist=False,
show_rug=False,
curve_type='kde')
# TBD: generalize this
if outputName == 'percent-change':
tickvalues = ['%d%%' % int(value) for value in barValues]
else:
tickvalues = ['%.2f' % value for value in barValues]
plotData = [
dict(type='bar',
x=barValues,
y=list(counts),
name=outputName,
histnorm='probability density',
marker=dict(color=colors),
showlegend=False,
text=tickvalues,
hoverinfo='text')
]
if 'kde' in distOptions:
distData = distPlot.data[0]
plotData.append(
dict(
type='scatter',
x=distData['x'],
y=distData['y'],
marker=dict(color=getColor('KDE')),
hoverinfo='skip', # prevents hover events
showlegend=False,
))
showMean = 'mean' in distOptions
showMedian = 'median' in distOptions
if showMean or showMedian:
info = values.describe()
maxY = max(counts)
labelSize = 12
bgcolor = getColor('PlotBg')
mean = info['mean']
median = info['50%']
def lineData(name, x, maxY, color):
d = dict(type='scatter',
mode='lines',
name=name,
x=[x, x],
y=[0.0000001, maxY],
marker=dict(color=color))
return d
def notation(x, y, xanchor, color):
# TBD: generalize this
if outputName == 'percent-change':
text = '%d%%' % x
else:
text = '%.2f' % x
d = dict(x=x,
y=y,
text=text,
xanchor=xanchor,
yanchor='top',
showarrow=False,
bgcolor=bgcolor,
opacity=0.7,
font=dict(family='Lato', size=labelSize, color=color))
return d
if showMean:
color = getColor('MeanLineColor')
xanchor = 'right' if mean <= median else 'left'
plotData.append(lineData('mean', mean, maxY, color))
annotations.append(notation(mean, 0.95 * maxY, xanchor, color))
if showMedian:
color = getColor('MedianLineColor')
xanchor = 'right' if median < mean else 'left'
plotData.append(lineData('median', median, maxY, color))
annotations.append(notation(median, maxY, xanchor, color))
return plotData, title, annotations