This project contains the code for "The contributions of phonological awareness and decoding on spelling in isiXhosa third grade readers". The first manuscript [EdArxiv link] has a pre-registration. However, owing to feedback from an initial round of review, we departed from the preregistration in the following ways in the revision to the manuscript [EdArxiv link]: * The original analysis included only start of Grade 3 data. One reviewer commented that to answer the research question we need longitudinal data. Data from the end of Grade 3 is now also included so there is a longitudinal component. The end of Grade 3 analysis in Study 2 is now directly comparable to study 1. * * Participants in study 2 were sampled from five schools. The clustering of participants in schools is now addressed through the use of linear mixed models, using lme4, and linear mixed quantile models, using lqmm. The use of mixed models produces less biased standard errors as it takes into account the role of clustering at school level. * * There is some missing data either at T1 or T2 in study 2. Missing data is now imputed using predictive mean matching from the mice package rather than using listwise deletion. This should result in less biased data since missing participants had worse spelling scores. The raw data is available on request from the corresponding author.