Exploring exploring student outcomes
David Kernohan is Deputy Editor of Wonkhe
Tags
For those of you ploughing your way through the workbook sent to your (OfS registered) provider this week, I have pulled the parallel Exploring Student Outcomes release together into a dashboard for ease of reference.
Here’s the main “Exploring Student Outcomes: Estimated differences” data. You can choose an “outcome” (completion, continuation, progression) and characteristic of interest and investigate how variations on the OfS modelling process differ from actual values – blue is the modelled difference (as chosen via the model factors drop down), and orange is the actual observed difference, from the reference category (shown in the title).
The bar chart below shows actual positive outcomes rates and the outcomes rates of the reference value. Right at the bottom you can adjust the population used (domicile, level of study, mode of study).
OfS tells us these values are generated using a combination of annual datasets – four years for most (longer for part time completion), but two years for progression (because we only have two years of Graduate Outcomes data). The OfS publication about this Exploring Student Outcomes data release, and an OfS dashboard, are also available.
For instances where the modelled values differ from the actual values – it appears that reality is at fault. No, seriously, that’s what the publication says:
If the bar gets smaller, this suggests the ‘actual difference’ is overstating the extent of
the relationship between the characteristic and the outcome … If the bar gets larger, this suggests the ‘actual difference’ is understating the extent of the relationship between the characteristic and the outcome.
Bonus data
If you download the dataset, you also get a “fixed effects estimates” sheet that doesn’t have an offical OfS dashboard. I’ve built you a Wonkhe one because of course I have.
What you are looking at is, on the face of it, quite similar to the main dashboard – but with a lot more options to examine coefficients and actual values for splits, and without the choice of models (everything on this one is the “fully saturated” variant). For the non-specialist the bar chart at the bottom is more interesting than the coefficients – you can see outcomes by top level subject, entry qualifications, household residual income, sexual orientation… all kinds of stuff. The width of each bar represents student numbers.