Effect size: commentary to the study on the Factors associated with the practice of exclusive breastfeeding

Abstract

This letter focuses on recent and interesting work on breastfeeding, to emphasize two observations. The first observation refers to the fact that, in Mateus and Cabrera’s manuscript  It was hardly discussed whether the referred knowledge and skills may be relevant to understanding the mothers’ behavior regarding their commitment to breastfeeding. The relevance of these cognitive aspects requires more attention due to their relationship with breastfeeding practices, and in general with the long-term mother-infant dyad.

Because the knowledge and skills to maintain successful breastfeeding have implications for developing instructional content in interventions for mothers as well, great attention needs to be paid to the size of the effect of differences between reported frequencies in pregnancy and the immediate puerperium. In Table 3, these differences were examined by the McNemar statistical test, which allows obtaining the statistical significance of the rejection of the null hypothesis of no differences. But neither this test nor the size of its p-value inform about the degree or size of the differences. An estimate of the size or magnitude of the differences, represented as point values or confidence intervals (as reported in Tables 4 and 5), tend to better specify tests of statistical significance.

Therefore, we estimate the practical significance of the difference between the percentages obtained in pregnancy and the immediate puerperium for each of the knowledge and basic skills reported in Table 3. With the only numerical data presented in this table, we calculated the McNemar odd ratio (McNemar OR , and standardized difference measures d: d Cox 8 and d probit . These show less bias with their population values in relation to other estimators5. Because they both assume different statistical assumptions e.g., logistic distribution for McNemar OR and d Cox , and normal distribution for d probit , calculating both
will report the convergence or divergence of these estimates.