Table 1.
Markov model transition probability matrix.
Table 2.
Validating graduation rates from the reduced-form Markov model.
Fig 1.
Using a Markov model to estimate SYGR (bottom panel) leads to tighter 95% confidence intervals than when a traditional approach is used (top panel). Shown are kernel density estimates for six-year graduation rate for the Fall 2013 (light gray), Fall 2014 (gray), and Fall 2015 (dark gray) cohorts and all three of these cohorts combined (orange).
Table 3.
Graduation rate estimates and confidence intervals.
Table 4.
Graduation rates and confidence intervals for small subgroups.
Fig 2.
Six-year graduation rate estimates for small subgroups of students.
Using a Markov model (bottom panels) to estimate SYGR for AALANA (left panels) and first-generation (right panels) science majors leads to tighter 95% confidence intervals and decreases year-to-year variability as compared to a traditional approach (top panels). Shown are kernel density estimates for six-year graduation rate for the Fall 2013 (light gray), Fall 2014 (gray), and Fall 2015 (dark gray) cohorts and all three of these cohorts combined (orange).
Fig 3.
Improvement in graduation rate from learning assistant support.
Learning Assistant support is associated with a 9 percentage point improvement in six-year graduation rates for science majors in general (a). These gains are even larger for AALANA science majors (b; 21 percentage point increase in SYGR) and for first-generation college students with science majors (c; 18 percentage point increase in SYGR).