Effects of treatment on nonschooled subjects

Unschooled subjects - that is, those for whom no schooling records were available and who were illiterate - were excluded from the main effects analyses and were analyzed separately, controlling only for socioeconomic differences and treatment x SES interactions.

Regressions were calculated for three of the psychoeducational tests (excluding literacy, reading, and the two Interamerican achievement tests since these required literacy skills) and the seven information-processing tasks. None of these analyses yielded a main effect of treatment at conventional levels of statistical significance or significant interactive terms.

As a result of the strict criteria imposed to define this group, sample sizes were quite small (ranging from 53 to 82). Power calculations for these analyses indicate relatively low power (around 58) for an effect size of .25 (Cohen, 1988), thus imposing a restriction on the interpretation of these results. Therefore, although no effects of Atole were observed within this restricted group, issues of low power preclude us from making definitive statements regarding the absence of effects in the unschooled sample.