Outcome & Treatment

Outcome

Control value is missing from data in design sample
Treatment is considered binary: either equal to the control value (control group) or not (treated group)
All values except the control value are considered as a different treatment

Estimated Average Treatment Effect

Assuming full randomization of the treatment variable (
{{ mlTaskDesign.treatmentVariable }}
), the Average Treatment Effect (ATE) can be estimated by the difference of the {{ isCausalClassification() ? 'frequency of the preferred class' : 'average outcome' }}
{{ mlTaskDesign.targetVariable }}
with treatment {{ mlTaskDesign.treatmentVariable }} is not {{ displayControlValue() }} and without. The estimated ATE is {{ outcomeStats.averageOutcomes.treated - outcomeStats.averageOutcomes.control | nicePrecision: 5 }}.

Estimated Average Treatment Effect

Assuming full randomization of the treatment variable (
{{ mlTaskDesign.treatmentVariable }}
), the Average Treatment Effect (ATE) for each treatment can be estimated by the difference of the {{ isCausalClassification() ? 'frequency of the preferred class' : 'average outcome' }}
{{ mlTaskDesign.targetVariable }}
with and without treatment.
The ATE estimated on the sample is negative{{ mlTaskDesign.enableMultiTreatment ? " across treatments" : "" }}, meaning {{ mlTaskDesign.enableMultiTreatment ? "treatments seem" : "the treatment seems" }} to have on average a negative effect on the outcome.
If this is unexpected, this could be due to a misconfiguration of the control group{{ isCausalClassification() ? " or the preferred class" : "" }}.
The values below are computed on a sample using the current sampling settings. You can edit them from the analysis script.
Outcome {{ isCausalClassification() ? 'value' : 'range' }}
Treatment
 {{ bin }} All
ATE
{{ isCausalClassification() ? 'Preferred class frequency' : 'Average outcome' }}
Control
{{ mlTaskDesign.treatmentVariable }}is {{ displayControlValue() }}
{{ outcomeStats.rowCountsByOutcomeBins[$index].control }} {{ outcomeStats.totalRowCounts.control }}
-
{{ outcomeStats.averageOutcomes.control | nicePrecision: 5 }}
Treatment
{{ mlTaskDesign.treatmentVariable }}is not {{ displayControlValue() }}
{{ outcomeStats.rowCountsByOutcomeBins[$index].treated | ifEmpty : '0' }} {{ outcomeStats.totalRowCounts.treated }}
{{ outcomeStats.averageOutcomes.treated | nicePrecision: 5 }}
Treatment: {{ displayTreatmentValue(treatment) }}
{{ mlTaskDesign.treatmentVariable }}is {{ displayTreatmentValue(treatment) }}
{{ outcomeStats.rowCountsByOutcomeBins[$index].perTreatment[treatment] | ifEmpty : '0' }} {{ outcomeStats.totalRowCounts.perTreatment[treatment] }}
{{ outcomeStats.averageOutcomes.perTreatment[treatment] - outcomeStats.averageOutcomes.control | nicePrecision: 5 }}
{{ outcomeStats.averageOutcomes.perTreatment[treatment] | nicePrecision: 5 }}
All {{ outcomeStats.rowCountsByOutcomeBins[$index].control + outcomeStats.rowCountsByOutcomeBins[$index].treated | ifEmpty : '0' }} {{ outcomeStats.totalRowCounts.control + outcomeStats.totalRowCounts.treated }}
{{ outcomeStats.averageOutcomes.treated - outcomeStats.averageOutcomes.control | nicePrecision: 5 }}
{{ outcomeStats.averageOutcomes.total | nicePrecision: 5 }}