The dataset contains all raw data used for analyses in the research of Study 2 described by Walda et al. (under review). The manuscript addressed the identification of determinants of dyslexia by traditional, linear techniques as well as machine learning techniques. For a full description of the reading and spelling remediation program, and all measures see the README file. The experiment design was a longitudinal repeated measures design. During two moments of measurement tests were administered in an one-to-one assessment setting. The pretest (T1) took place prior to all interventions, and the follow-up measurements was administered after three months (T2) of reading and spelling remediation. In this document, the process of collecting data is described, followed by an overview and description of variables included in the data set. The data set is based on an earlier version, used for analyses in the research of Walda, van Weerdenburg, Wijnants, and Bosman (2015). The earlier version of the dataset will be published and linked. In the present dataset, both variables and cases have been added. However, the present data set includes only two moments of measurement, instead of four. The dataset includes test scores for word decoding, grapheme-phoneme identification, grapheme-phoneme discrimination, naming speed, vocabulary, nonverbal reasoning, digit recall, block recall, and word recall at T1 in children with severe dyslexia prior to a reading and spelling remediation program in a specialized reading clinic in the Netherlands. The dataset includes test scores for word decoding at T2, after three months of reading and spelling remediation.