def plot_gender_distributions_over_years(data, title='Gender distribution over the years for the different runnings of Lausanne Marathon'):
'''
This function generates a graph representing gender distributions over the years, given a data set.
Parameters
- data: Array containing all DataFrames to consider (one by gender)
- title: Title of the graph (by default, 'Gender distribution over the years for different runnings of Lausanne Marathon')
Return
- figure: Plotly figure
'''
colors = {'10 km': KM_10_COLOR, 'Semi-marathon': KM_21_COLOR, 'Marathon': KM_42_COLOR}
# We ignore outputs of Plotly
with study_utils.ignore_stdout():
figure = tools.make_subplots(rows=1, cols=2, subplot_titles=([(sex.capitalize() + ' runners') for sex in data]))
i = 1
for sex, df in data.items():
bars = []
for running in df.columns:
figure.append_trace(go.Bar(x=[year for year in df.index], y=df[running], name=running, marker={'color': colors[running]}, legendgroup=running, showlegend=(i == 1)), 1, i)
i += 1
figure['layout'].update(title=title, barmode='stack')
plotly.offline.iplot(figure)
return figure
def plot_performance_comparison(data, age_category, year, performance_criterion, silent=False):
'''
This function displays performance comparison graph given a set of data and a performance criterion.
Parameters
- data: Dict containing all the data used for performance comparison (see generate_performance_comparison)
- age_category: string representing age category considered
- year: year to compare with all other Lausanne Marathon editions
- performance_criterion: Criterion to use for y axis
- silent: If True, outputs are not displayed and figure is only returned
'''
criterion = performance_criterion.lower()
# Creation of figure (output is ignored)
with study_utils.ignore_stdout():
figure = tools.make_subplots(rows=1, cols=3, subplot_titles=([key for key, value in data[criterion].items()]))
for index, boxplots in enumerate(data[criterion]):
for boxplot in data[criterion][boxplots]:
figure.append_trace(boxplot, 1, index+1)
if criterion == 'time':
for axis, attributes in {k: v for k, v in figure['layout'].items() if 'yaxis' in k}.items():
figure['layout'][axis].update(type='date', tickformat='%H:%M:%S')
figure['layout']['yaxis1'].update(title=performance_criterion)
figure['layout'].update(title='Performance comparison between Lausanne Marathon ' + str(year) + ' and all Lausanne Marathon<br>(Age category: ' + age_category +')')
if not silent:
plotly.offline.iplot(figure)
return figure
def generate_performance_distribution_figure(data, age_category, sex_category, criterion):
'''
This function generates performance distribution figure given a dataset.
Parameters
- data: DataFrame containing data to use
- age_category: Selected age category
- sex_category: Selected sex category
- criterion: Selected criterion
Return
- figure: Plotly figure
'''
# We define the considered runnings and the years interval
runnings = OrderedDict([(10, {'name': '10 km', 'position': 1}), (21, {'name': 'Semi-marathon', 'position': 2}), (42, {'name': 'Marathon', 'position': 3})])
years_range = range(1999, 2017)
years = {year: str(year) for year in years_range}
colors = study_utils.generate_colors_palette(years)
# We ignore outputs of Plotly
with study_utils.ignore_stdout():
figure = tools.make_subplots(rows=3, subplot_titles=([attributes['name'] for km, attributes in runnings.items()]))
for km, attributes in runnings.items():
for year in years_range:
data_filtered = data[(data['distance (km)'] == km) & (data['year'] == year)]
if criterion == 'time':
group_data_filtered = data_filtered.set_index('time').groupby(pd.TimeGrouper(freq='5Min'))
x_values = [datetime.datetime.strptime(str(name), '%Y-%m-%d %H:%M:%S') for name, group in group_data_filtered]
elif criterion == 'speed (m/s)':
group_data_filtered = data_filtered.round({'speed (m/s)': 1}).groupby('speed (m/s)')
x_values = [name for name, group in group_data_filtered]
line = go.Scatter(mode='lines', x=x_values, y=[len(group) for name, group in group_data_filtered], name=str(year), legendgroup=str(year), marker={'color': colors[year]}, showlegend=(attributes['position'] == 1))
figure.append_trace(line, attributes['position'], 1)
# Format of x axes if time is selected criterion
if criterion == 'time':
for axis, attributes in {k: v for k, v in figure['layout'].items() if 'xaxis' in k}.items():
figure['layout'][axis].update(type='date', tickformat='%H:%M')
figure['layout'].update(width=1000, height=600, margin=go.Margin(t=100, b=50, l=50, r=50))
figure['layout'].update(title='Performance distribution of Lausanne Marathon editions for all runnings<br>(Age category: ' + age_category +' | Sex category: ' + sex_category +')')
figure['layout']['legend'].update(y=0.5)
return figure
def generate_evolution_and_performance_figure(data, criterion):
'''
This function generates single evolution and performance figure given a set of data (see generate_evolution_and_performance_figures).
Parameters
- data: Data to be used for generation of figure
- criterion: Performance criterion to be used
Return
- figure: Plotly figure representing evolution and performance in same graph
'''
year_values = [year for year in range(1999, 2017)]
year_labels = [v for v in year_values]
colors = {'10 km': KM_10_COLOR, 'Semi-marathon': KM_21_COLOR, 'Marathon': KM_42_COLOR}
# We ignore outputs of Plotly
with study_utils.ignore_stdout():
figure = tools.make_subplots(rows=4, specs=[[{'rowspan': 2}], [None], [{'rowspan': 2}], [None]], subplot_titles=('Evolution of runners over the years', 'Performance comparison by years'))
# Add all lines in first subplot
for line in data['evolution']:
line['legendgroup'] = line['name']
line['marker'] = {'color': colors[line['name']]}
figure.append_trace(line, 1, 1)
# Add all boxplots in second subplot
for boxplot in data['performance'][criterion]:
boxplot['legendgroup'] = boxplot['name']
boxplot['marker'] = {'color': colors[boxplot['name']]}
boxplot['name'] = ''
figure.append_trace(boxplot, 3, 1)
# Format of x axes
for axis, attributes in {k: v for k, v in figure['layout'].items() if 'xaxis' in k}.items():
figure['layout'][axis].update(tickvals=year_values, ticktext=year_labels)
# Format of y axes
figure['layout']['yaxis1'].update(title='Number of runners', tickformat='f')
figure['layout']['yaxis2'].update(title=criterion)
if criterion.lower() == 'time':
figure['layout']['yaxis2'].update(type='date', tickformat='%H:%M')
# Use of group for boxplots to avoid superposition
figure['layout'].update(boxmode='group')
# Update of size, margins and legend attributes of figure
figure['layout'].update(width=1000, height=600, margin=go.Margin(t=50, b=50, l=50, r=50))
figure['layout']['legend'].update(y=0.5)
return figure
def plot_ols_fitted_and_true_values(data, ols_results, x=None, y='Median age (all runnings)', title='Fitted and original values for median ages of participants of Lausanne Marathon editions', groupby_column='Gender', markers_attributes=None):
'''
This function plots fitted and true values for a given dataset and associated ols results.
Parameters
- data: DataFrame containing data to use for original values
- x: Name of the column to use for x axis for original values and fitted values (by default None / if None, index will be used for original values and index name for retrieve_ols_predictions_and_errors)
- y: Name of the column to use for y axis (by default, 'Median age (all runnings)')
- title: Title of the graph (by default, 'Fitted and original values for median ages of participants of Lausanne Marathon editions')
- groupby_column: Name of the column to use for grouping data (by default, 'Gender')
- markers_attributes: Dictionary containing options for each unique value in column groupby_column (by default, None)
(at present, 'title' (title of subplot), 'color_true' (color of markers of original values), 'color_fitted' (color of markers of fitted values), 'color_errors' (color of errors bars), 'name_true' (name of markers' legend of original values) and 'name_fitted' (name of of markers' legend of fitted values) are supported)
Return
- figure: Plotly figure
'''
position = {}
predictions_and_errors = study_utils.retrieve_ols_predictions_and_errors(ols_results=ols_results, regressor=(x if x else data.index.name))
# We ignore outputs of Plotly
with study_utils.ignore_stdout():
figure = tools.make_subplots(rows=1, cols=2, shared_yaxes=True, subplot_titles=([(markers_attributes[key].get('title', None) if markers_attributes else None) for key in predictions_and_errors]))
i = 1
for key, results in predictions_and_errors.items():
position[key] = i
name=(markers_attributes[key].get('name_fitted', 'Fitted values') if markers_attributes else 'Fitted values')
errors = {'type': 'data', 'symmetric': False, 'array': np.subtract(np.array(results['errors']['max']), np.array(results['predictions'])), 'arrayminus': abs(np.subtract(np.array(results['errors']['min']), np.array(results['predictions']))), 'color': (markers_attributes[key].get('color_errors', None) if markers_attributes else None)}
markers = go.Scatter(x=results['x'], y=results['predictions'], mode='markers', hoverinfo='y+text', name=name, text=name, marker={'color': (markers_attributes[key].get('color_fitted', None) if markers_attributes else None)}, legendgroup='fitted', error_y=errors, showlegend=(position[key] == 1))
figure.append_trace(markers, 1, position[key])
i += 1
for key, group in data.groupby([groupby_column]):
x_values = group[x] if x else group.index
name = (markers_attributes[key].get('name_true', 'Original values (' + key + ')') if markers_attributes else 'Original values (' + key + ')')
markers = go.Scatter(x=x_values, y=group[y], mode='markers', hoverinfo='y+text', name=name, text=name, marker={'color': (markers_attributes[key].get('color_true', None) if markers_attributes else None)})
figure.append_trace(markers, 1, position[key])
# Add of title and format of axes
figure['layout'].update(title=title)
figure['layout']['yaxis'].update(title=y)
for axis, attributes in {k: v for k, v in figure['layout'].items() if 'xaxis' in k}.items():
figure['layout'][axis].update(title=(x if x else data.index.name))
plotly.offline.iplot(figure)
return figure
def generate_comparison(data, title, running_type_column_name, x_column_name,
y_column_name, colors):
'''
This function generates comparison boxplots according to given parameters, for each running of Lausanne Marathon.
Parameters
- data: DataFrame containing information on runners
- title: Title of the graph
- running_type_column_name: Name of the column containing the runnings
- x_column_name: Name of column to be used for x data
- y_column_name: Name of column to be used for y data
- colors: Dict containing colors associated to each unique value of x_column_name
Return
- figure: Plotly figure
'''
runnings = OrderedDict([(10, {
'name': '10 km',
'position': 1
}), (21, {
'name': 'Semi-marathon',
'position': 2
}), (42, {
'name': 'Marathon',
'position': 3
})])
# We create final figure (outputs are ignored)
with study_utils.ignore_stdout():
figure = tools.make_subplots(
rows=1,
cols=3,
shared_yaxes=True,
subplot_titles=([
attributes['name'] for km, attributes in runnings.items()
]))
# Loop over runnings of Lausanne Marathon
for running, attributes in runnings.items():
filtered_df = data[data[running_type_column_name] == running]
# We loop over gender of participants
for value in np.concatenate(
(['all'], filtered_df[x_column_name].unique()), axis=0):
if value == 'all':
filtered_df_final = filtered_df
else:
filtered_df_final = filtered_df[filtered_df[x_column_name] ==
value]
# We append a new boxplot corresponding to a gender (or all participants) for the considered running
figure.append_trace(
go.Box(y=filtered_df_final[y_column_name],
name=value.capitalize() + ' runners',
marker={'color': colors[value]},
legendgroup=value,
showlegend=(attributes['position'] == 1)), 1,
attributes['position'])
# Format of y and x axes and modification of layout
for axis, _ in {k: v
for k, v in figure['layout'].items()
if 'xaxis' in k}.items():
figure['layout'][axis].update(showticklabels=False)
figure['layout']['yaxis1'].update(title=y_column_name.capitalize(),
tickformat='.2f')
figure['layout'].update(title=title)
plotly.offline.iplot(figure)
return figure
def plot_speed_distribution_by_running(data,
runnings=None,
title='Speed distribution by running',
speed_column_name='speed (m/s)',
sex_column_name='sex'):
'''
This function plots, for each running, the distribution of ages of runners based on the genders of participants.
Parameters
- data: DataFrame to use during generation of the distribution
- runnings: Dict containing name of column containing runnings (key: column_name) and set of runnings (key: values, value: dict() with following keys: name, color)
By default, None. If None, default values will be set by function.
- title: Title of the graph (by default, 'Speed distribution by running')
- speed_column_name: Name of the column containing age of participants('age' or 'age category', by default, 'speed (m/s)')
- sex_column_name: Name of the column containing sex of participants (by default, 'sex')
Return
- figure: Plotly figure
title, x axis name, statistics, column to select, categories or not, title of xaxis, title of yaxis (not modified)
'''
if not runnings:
runnings = {
'column_name':
'distance (km)',
'values':
OrderedDict([(10, {
'name': '10 km',
'color': KM_10_COLOR,
'position': 1
}),
(21, {
'name': 'Semi-marathon',
'color': KM_21_COLOR,
'position': 2
}),
(42, {
'name': 'Marathon',
'color': KM_42_COLOR,
'position': 3
})])
}
colors = {
'female': FEMALE_COLOR,
'male': MALE_COLOR,
'all': ALL_GENDERS_COLOR
}
statistics = {}
with study_utils.ignore_stdout():
figure = tools.make_subplots(
rows=3,
cols=1,
shared_xaxes=True,
subplot_titles=([
attributes['name']
for km, attributes in runnings['values'].items()
]))
for key, attributes in runnings['values'].items():
filtered_df = data[data[runnings['column_name']] == key]
statistics[attributes['name']] = 'Total: ' + str(
len(filtered_df)) + ' runners<br>Max: ' + str(
round(np.max(filtered_df[speed_column_name]),
2)) + ' m/s<br>Min: ' + str(
round(np.min(filtered_df[speed_column_name]), 2)
) + ' m/s<br>Median: ' + str(
round(np.median(filtered_df[speed_column_name]),
2)) + ' m/s | SD: ' + str(
round(
np.std(filtered_df[speed_column_name]),
2)) + ' m/s'
for sex in np.concatenate(
(filtered_df[sex_column_name].unique(), ['all']), axis=0):
if sex == 'all':
x = filtered_df[speed_column_name]
else:
x = filtered_df[filtered_df[sex_column_name] ==
sex][speed_column_name]
figure.append_trace(
go.Histogram(xbins={
'start':
math.floor(np.min(data[speed_column_name])),
'end':
math.ceil(np.max(data[speed_column_name])),
'size':
0.1
},
x=x,
name=sex.capitalize() + ' runners',
legendgroup=sex,
showlegend=(attributes['position'] == 1),
marker={'color': colors[sex]},
opacity=0.75), attributes['position'], 1)
# Format of axes and layout
figure.layout.xaxis1.update(title='Speed (m/s)', tickformat='.1f')
figure.layout.yaxis2.update(title='Number of participants')
figure.layout.update(title=title,
barmode='stack',
bargroupgap=0.1,
bargap=0,
margin=go.Margin(t=100, b=50, l=50, r=50))
# Add of statistics
# Trick: We use position of subtitles annotations to create the ones related to statistics
annotations_statistics = []
for annotation in figure['layout']['annotations']:
annotations_statistics.append(
Annotation(y=annotation['y'] - 0.12,
x=1,
align='left',
text=statistics[annotation['text']],
xref='paper',
yref='paper',
yanchor='bottom',
showarrow=False))
figure['layout']['annotations'].extend(annotations_statistics)
plotly.offline.iplot(figure)
return figure
def generate_performance_by_age_and_age_category(
data,
runnings=None,
age_column_name='age',
age_category_column_name='age category',
sex_column_name='sex'):
'''
This function generates figures for each running. It displays time distribution according to age and age category.
Final Dict has the following pattern:
{
<running_1>: <Plotly figure>
[, <running_2>: <Plotly figure>
, ...]
}
Parameters
- data: DataFrame containing all the information of Lausanne Marathon 2016
- runnings: Dict containing name of column containing runnings (key: column_name) and set of runnings (key: values, value: dict() with following keys: name)
By default, None. If None, default values will be set by function.
- age_column_name: Name of the column containing age of participants(by default, 'age')
- age_category_column_name: Name of the column containing age category of participants(by default, 'age category')
- sex_column_name: Name of the column containing gender of participants (by default, 'sex')
Return
- figures: Dict containing all time distribution figures
'''
# We create final dict and we set attributes and runnings (if not given by user)
figures = {}
attributes = {
'colors': {
'female': FEMALE_COLOR,
'male': MALE_COLOR,
'all': ALL_GENDERS_COLOR
},
'names': {
'female': 'Female runners',
'male': 'Male runners',
'all': 'All runners'
},
'visibility': {
'female': 'legendonly',
'male': 'legendonly',
'all': True
}
}
if not runnings:
runnings = {
'column_name': 'distance (km)',
'values': {
10: {
'name': '10 km'
},
21: {
'name': 'Semi-marathon'
},
42: {
'name': 'Marathon'
}
}
}
# Loop over runnings
for km, attributes_running in runnings['values'].items():
filtered_df = data[data[runnings['column_name']] == km]
# Creation of Plotly figure containing subplots (we ignore outputs here)
with study_utils.ignore_stdout():
figure = tools.make_subplots(rows=2, cols=1, vertical_spacing=0.1)
# Consideration of ages and age categories
for column_name in [age_column_name, age_category_column_name]:
# Creation of boxplots for female and male runners
boxplots = study_utils.create_plotly_boxplots(
data=filtered_df,
x=column_name,
y='time',
hue=sex_column_name,
hue_names=attributes['names'],
colors=attributes['colors'],
visibility=attributes['visibility'],
use_hue_names=False,
use_legend_group=True,
show_legend=(column_name == 'age'))
# We add boxplots for all runners (without consideration of sex)
boxplots.append(
go.Box(y=filtered_df['time'],
x=filtered_df[column_name],
name=attributes['names']['all'],
marker={'color': attributes['colors']['all']},
visible=attributes['visibility']['all'],
legendgroup=attributes['names']['all'],
showlegend=(column_name == 'age')))
# We add each generated boxplot in the correct subplot
for boxplot in boxplots:
figure.append_trace(boxplot, 1 if column_name == 'age' else 2,
1)
# Format of y and x axes and modification of layout
for axis, _ in {
k: v
for k, v in figure['layout'].items() if 'yaxis' in k
}.items():
figure['layout'][axis].update(title='Performance time',
type='date',
tickformat='%H:%M:%S')
figure['layout']['xaxis1'].update(title='Age', dtick=5)
figure['layout']['xaxis2'].update(title='Age category',
categoryorder='array',
categoryarray=YEAR_CATEGORIES)
figure['layout'].update(title='Distribution of time performance (' +
attributes_running['name'] + ')',
height=650,
hovermode='closest')
# We add newly created figure to the final dict
figures[attributes_running['name']] = figure
return figures
def plot_distribution_age_distance(data,
runnings=None,
title='Distribution of runners by age',
age_column_name='age',
sex_column_name='sex'):
'''
This function plots, for each running, the distribution of ages of runners based on the genders of participants.
Parameters
- data: DataFrame to use during generation of the distribution
- runnings: Dict containing name of column containing runnings (key: column_name) and set of runnings (key: values, value: dict() with following keys: name, color)
By default, None. If None, default values will be set by function.
- title: Title of the graph (by default, 'Distribution of runners by age categories')
- age_column_name: Name of the column containing age of participants('age' or 'age category', by default, 'age')
- sex_column_name: Name of the column containing sex of participants (by default, 'sex')
Return
- figure: Plotly figure
'''
if not runnings:
runnings = {
'column_name':
'distance (km)',
'values':
OrderedDict([(10, {
'name': '10 km',
'color': KM_10_COLOR,
'position': 1
}),
(21, {
'name': 'Semi-marathon',
'color': KM_21_COLOR,
'position': 2
}),
(42, {
'name': 'Marathon',
'color': KM_42_COLOR,
'position': 3
})])
}
colors = {
'female': FEMALE_COLOR,
'male': MALE_COLOR,
'all': ALL_GENDERS_COLOR
}
statistics = {}
with study_utils.ignore_stdout():
figure = tools.make_subplots(
rows=3,
cols=1,
subplot_titles=([
attributes['name']
for km, attributes in runnings['values'].items()
]))
for key, attributes in runnings['values'].items():
filtered_df = data[data[runnings['column_name']] == key]
statistics[attributes['name']] = 'Mean age: ' + str(
round(np.mean(filtered_df[age_column_name]),
2)) + ' years (SD: ' + str(
round(np.std(filtered_df[age_column_name]), 2)) + ')'
for sex in np.concatenate(
(filtered_df[sex_column_name].unique(), ['all']), axis=0):
if sex == 'all':
x = filtered_df[age_column_name]
else:
x = filtered_df[filtered_df[sex_column_name] ==
sex][age_column_name]
nbinsx = ((np.max(x) - np.min(x)) +
1) if (age_column_name == 'age') else len(x)
figure.append_trace(
go.Histogram(nbinsx=nbinsx,
x=x,
name=sex.capitalize() + ' runners',
legendgroup=sex,
showlegend=(attributes['position'] == 1),
marker={'color': colors[sex]},
opacity=0.75), attributes['position'], 1)
# Format of axes and layout
if age_column_name == 'age category':
for axis, attributes in {
k: v
for k, v in figure['layout'].items() if 'xaxis' in k
}.items():
figure['layout'][axis].update(categoryorder='array',
categoryarray=YEAR_CATEGORIES)
figure.layout.xaxis3.update(title='Age of participants')
figure.layout.yaxis2.update(title='Number of participants')
figure.layout.update(title=title,
barmode='stack',
bargroupgap=0.1,
bargap=0,
margin=go.Margin(t=100, b=50, l=50, r=50))
# Add of statistics
# Trick: We use position of subtitles annotations to create the ones related to statistics
annotations_statistics = []
for annotation in figure['layout']['annotations']:
annotations_statistics.append(
Annotation(y=annotation['y'],
x=1,
text=statistics[annotation['text']],
xref='paper',
yref='paper',
yanchor='bottom',
showarrow=False))
figure['layout']['annotations'].extend(annotations_statistics)
plotly.offline.iplot(figure)
return figure