Learning progressions rising field, in science ed

Little work on measurement/psychometric modelling of LP data

Many choices in how to model data

Insights for scholars who want to use Bayesian networks to examine models with respect to parameter recovery

Center for continuous intructional improvement

A testable hypothesis regarding how students' understanding and ability grows over time with appropriate instruction

• learning target or clear end point
• progress variables
• defined significant stages
• operational definitions of each of the stages
• assessments

Graphical representation of structure - relationships between variables (vertices) - observable or latent - edges : probabilistic dependency.

Probability of a variable is conditional on it's parent's nodes.

Students might take different paths. How will model reflect learning progression? This study: 4 structures, all variations of modelling a single LP. Will generate data under all structures, and analyze using each of the structures.

Instead of it being a progression, different stages that students' are at. You can have some level 3 skills, some level 2 skills, etc.

Or maybe there is some relationship or dependency between the levels (probability, not deterministic). You can still jump around, not necessarily go in order.

Dependency of having level 4 skills, might depend on having all the other levels (complex). Maybe skill levels are not in line, but parallell?

Factors to vary:

• sample size
• distrib of students in latent classes

Keep constant

• no of levels of learning progression
• no of observable variables
• strenght of relationships between observable variables and latent variables

Not much benefit to separating out levels of learning progression

In this case skills were still related - look at a case where there is more of a separation between kills