I know Marchesman has previously tried to use the Samoylova and Hall paper to show that the state education effect is as significant as that of cognitive and learning disabilities.
Worth noting that when you read the Samoylova and Hall paper it states:
The results show that taken individually, all of the factors remain fairly poor predictors of the outcome. For example, the top ranked predictor based on R-squared value for the first class outcome variable only explains 5.3% of the variance in that outcome.
Across all three outcomes, the results obtained in the first year of study and the course of study (Tripos) remain the two strongest predictors in that order.
And:
It is important to note that the statistically significant result for a particular category does not indicate the importance or causal effect on the outcomes - as highlighted above, the predictive powers of all the factors on the outcome remain very low.
Interestingly, although Marchesman only showed the part of the table referring to the multivariate model for first class degrees, when the results for good honours degrees (first class and upper second) are shown, comprehensive schools are no longer statistically significant.
The paper concludes:
Both univariate and multivariate modelling of a number of quantitative factors indicate a relatively low predictive strength of these factors, both as individual predictors and as a combined set. It is not possible to numerically capture and model all characteristics of an individual and use them in a predictive model. It can be hypothesised that a number of attributes related to teaching and learning approaches, such as study and revision behaviours, levels of self-confidence, peer-group interactions, content of education activities and assessment types, which are intrinsically harder to measure quantitatively and which thus remained outside of the scope of this investigation, can be influencing the outcomes.
Of all factors studied, the attainment in the first year of the course was consistently shown to be the best predictor amongst the rest, which is not surprising as previous attainment is a known indicator of future academic performance.
https://www.cao.cam.ac.uk/files/attainment_outcomes.pdf