Spelling–sound conflict cost did not reduce observed mean squared word-reading prediction error in fixed ridge models with age and mean accuracy for 90 children

We tested whether spelling–sound conflict cost adds information about word-reading skill. We used 90 child records from OpenNeuro ds006239, release 1.0.7. Each child had two word-rhyming runs and a letter-word identification standard score. We counted missed responses as errors. The fixed ridge models used age and mean accuracy. We excluded each tested child from model fitting. Adding conflict cost changed mean squared error from 98.81 to 99.35 score points squared. The error reduction was -0.54. Its approximate 95 percent bootstrap interval was -13.83 to 10.75. This interval permits both benefit and harm. The run correlation was 0.283 for conflict cost and 0.773 for mean accuracy. Model checks gave mixed error changes. The package preserves source files, hashes, the local analysis plan, code, predictions, and checks. These results concern this sample and these models. They do not establish that conflict cost has no predictive value.

word reading · phonology · spelling-sound conflict · prediction · ridge regression · individual differences · children · behavioral task