This article is only available in the PDF format. Download the PDF to view the article, as well as its associated figures and tables.
Sir.—It is difficult to disagree with the interest shown by Dr Goldson et al (Am J Dis Child 132:790-793, 1978) in the promotion of "good parenting skills." However, their failure to discuss bias and their confusing use of graphs result in a paper that is less than persuasive. If chance does not explain an observed association in a sample, it does not necessarily follow that the observed association is valid. There are several forms of bias that may produce an observed association in a sample that is not valid for the referent population.
The article by Goldson et al demonstrates at least two forms of bias that deserve consideration. The first is selection bias. How do their 52 abused and 23 nonabused study children compare with the 140 subjects in the initial prospective study and is it reasonable to generalize from this nonrandomly selected sample of abused children to