|
Field
Experimental Design (Public)
|
Before
DESIGN OVERVIEW
The study is a classroom-level cluster-randomized controlled trial, with randomization stratified by school.
The intervention is implemented among students enrolled in 2º ESO (second year of compulsory secondary education of the Spanish school system). Within each participating school, all classrooms corresponding to 2º ESO are identified prior to randomization.
Randomization occurs at the classroom level within each school and within this grade. Specifically, at least one 2º ESO classroom per school is randomly assigned to the treatment condition, while at least one other 2º ESO classroom in the same school serves as the control group.
This design is therefore not a school-level randomization, but a within-school, within-grade cluster-randomized design at the classroom level.
WITHIN-SCHOOL AND WITHIN-GRADE STRUCTURE
Because both treatment and control classrooms belong to the same grade (2º ESO) within the same schools, this design has two key implications. First, it enhances internal comparability, as treatment and control students share the same school leadership and infrastructure, the same grade-level curriculum, and the same academic calendar and evaluation system. This reduces confounding from school-level differences and increases comparability across study arms. Second, it introduces the potential for peer spillovers. Because treatment and control classrooms coexist within the same grade cohort, students may interact across classrooms, potentially generating spillover effects if treated students share workshop content or attitudes with control students. Such spillovers, if present, would likely attenuate estimated treatment effects toward zero. The primary estimand therefore remains the intention-to-treat effect of classroom-level assignment, interpreted under potential partial interference across classrooms.
SAMPLE STRUCTURE
We collect our sample using the following procedure. We first run a pilot experiment (Phase I) within 1 public secondary school in Madrid. We randomize between 2º ESO classrooms (how many? ). We end up with a total of 81 students, among which 45 were assigned to treatment classroom and 36 were assigned to control classroom.
In a second phase (Phase 2), we fully implement our experiment in 11 secondary schools in Madrid (7 public and 4 semi-private/private). We end up with a sample of 751 students belonging to 11 treated classrooms and 23 control classrooms. Each school contributes at least two 2º ESO classrooms. In schools with two classrooms, one classroom is assigned to treatment and one to control. In schools with more than two classrooms in the grade, additional classrooms are randomly allocated while ensuring representation in both arms.
|
After
DESIGN OVERVIEW
The study is a classroom-level cluster-randomized controlled trial, with randomization stratified by school.
The intervention is implemented among students enrolled in 2º ESO (second year of compulsory secondary education of the Spanish school system). Within each participating school, all classrooms corresponding to 2º ESO are identified prior to randomization.
Randomization occurs at the classroom level within each school and within this grade. Specifically, at least one 2º ESO classroom per school is randomly assigned to the treatment condition, while at least one other 2º ESO classroom in the same school serves as the control group.
This design is therefore not a school-level randomization, but a within-school, within-grade cluster-randomized design at the classroom level.
WITHIN-SCHOOL AND WITHIN-GRADE STRUCTURE
Because both treatment and control classrooms belong to the same grade (2º ESO) within the same schools, this design has two key implications. First, it enhances internal comparability, as treatment and control students share the same school leadership and infrastructure, the same grade-level curriculum, and the same academic calendar and evaluation system. This reduces confounding from school-level differences and increases comparability across study arms. Second, it introduces the potential for peer spillovers. Because treatment and control classrooms coexist within the same grade cohort, students may interact across classrooms, potentially generating spillover effects if treated students share workshop content or attitudes with control students. Such spillovers, if present, would likely attenuate estimated treatment effects toward zero. The primary estimand therefore remains the intention-to-treat effect of classroom-level assignment, interpreted under potential partial interference across classrooms.
SAMPLE STRUCTURE
We collect our sample using the following procedure. We first run a pilot experiment (Phase I) within 1 public secondary school in Madrid. We randomize between 2º ESO classrooms. We end up with a total of 81 students, among which 45 were assigned to treatment classroom and 36 were assigned to control classroom.
In a second phase (Phase 2), we fully implement our experiment in 11 secondary schools in Madrid (7 public and 4 semi-private/private). We end up with a sample of 751 students belonging to 11 treated classrooms and 23 control classrooms. Each school contributes at least two 2º ESO classrooms. In schools with two classrooms, one classroom is assigned to treatment and one to control. In schools with more than two classrooms in the grade, additional classrooms are randomly allocated while ensuring representation in both arms.
|