Hi there,
My colleague and I are trying to evaluate the impact of a tutoring intervention and have test score data for treated children and a control group. However, study enrolment began earlier for the treatment group than the control group. Therefore, the treatment group could potentially have received the intervention for a maximum of three months prior to baseline measurement in the control group at enrolment.
There are two different tests:
Test 1 (imbalanced data): Available since implementation began for treatment group, but available only after enrolment for control group (i.e. earlier measurements available for treatment group)
Test 2 (pre-treatment exposure): For both groups, data are only available after enrolment began for the control group (i.e. datasets are balanced, but the treatment group had already received up to 3 months of intervention prior to first data point).
Are there any mitigation strategies in either case? For test 1, is there any way to account for earlier test score data for the treatment group only? For test 2, do I just need to accept that any coefficient for effect size will effectively be a lower bound estimate?
Many thanks in advance!
My colleague and I are trying to evaluate the impact of a tutoring intervention and have test score data for treated children and a control group. However, study enrolment began earlier for the treatment group than the control group. Therefore, the treatment group could potentially have received the intervention for a maximum of three months prior to baseline measurement in the control group at enrolment.
There are two different tests:
Test 1 (imbalanced data): Available since implementation began for treatment group, but available only after enrolment for control group (i.e. earlier measurements available for treatment group)
Test 2 (pre-treatment exposure): For both groups, data are only available after enrolment began for the control group (i.e. datasets are balanced, but the treatment group had already received up to 3 months of intervention prior to first data point).
Are there any mitigation strategies in either case? For test 1, is there any way to account for earlier test score data for the treatment group only? For test 2, do I just need to accept that any coefficient for effect size will effectively be a lower bound estimate?
Many thanks in advance!
