When analyzing your fitted Meridian model, the Analyzer, Visualizer, and
BudgetOptimizer classes allow one to scope the analysis to specific
geographical regions or time periods. These arguments allow you to calculate
metrics (like ROI or incremental outcome), generate visualizations, and perform
budget optimization based on subsets of your input data.
Use selected_geos argument for scoping to geos. Use selected_times argument
for scoping to time periods for the Analyzer and Visualizer classes. To
define the time window for BudgetOptimizer, use start_date and end_date
arguments (typically provided in yyyy-mm-dd string format).
By default, these are set to include all the geographical units and all of the
time periods respectively. The boolean options aggregate_times and
aggregate_geos available in some methods lets you to either segment results
be time and geo respectively, or to aggregate the selected times and geos.
Example code
Several hypothetical scenarios are presented in this section for how to use
selected_geos and selected_times.
Scenario 1: Calculate summary metrics across a specific time period
# Scenario 1: Calculate summary metrics for Q4 2024 only, aggregated across geos.
from meridian.analysis import analyzer
from meridian.model import model
mmm = model.Meridian(...)
mmm.sample_posterior(...)
analyzer_instance = analyzer.Analyzer(mmm)
selected_q4_times = [
'2024-10-01',
'2024-10-08',
'2024-10-15',
'2024-10-22',
'2024-10-29',
'2024-11-05',
'2024-11-12',
'2024-11-19',
'2024-11-26',
'2024-12-03',
'2024-12-10',
'2024-12-17',
'2024-12-24',
'2024-12-31',
]
q4_national_summary = analyzer_instance.summary_metrics(
selected_times=selected_q4_times,
)
Scenario 2: Calculate incremental outcome for 2 hypothetical geos
# Scenario 2: Calculate incremental outcome for 2 hypothetical geos, 'Geo_A', 'Geo_B' broken down weekly.
from meridian.analysis import analyzer
from meridian.model import model
mmm = model.Meridian(...)
mmm.sample_posterior(...)
analyzer_instance = analyzer.Analyzer(mmm)
# Note: Incremental Outcome can be aggregated by time or shown weekly.
geo_a_weekly_impact = analyzer_instance.incremental_outcome(
selected_geos=['Geo_A', 'Geo_B'],
aggregate_geos=False, # Keep the geo dimension separate.
aggregate_times=False, # Show incremental outcome week-by-week.
)
Scenario 3: Filtering Model Fit Plot
# Scenario 3: Filtering Model Fit Plot
# You can restrict the ModelFit.plot_model_fit() visualization to a specific
# region and ensure results are shown at that detailed geo level.
# This example assumes weekly aggregation.
from meridian.model import model
from meridian.analysis import visualizer
mmm = model.Meridian(...)
mmm.sample_posterior(...)
# Assume 'mmm' is fitted
model_fit = visualizer.ModelFit(mmm)
model_fit_chart = model_fit.plot_model_fit(
selected_geos=['Geo_A'],
# Plotting January 2025
selected_times=[
'2025-01-01',
'2025-01-08',
'2025-01-15',
'2025-01-22',
'2025-01-29',
],
# Display Geo_A's plot separately (relevant for multi-geo models)
show_geo_level=True,
include_baseline=True,
)