In order to analyze the different statistical analysis performed in human services peer-reviewed research articles, a paper by Anderson et al. (2005) has been chosen, where the researchers designed a randomized controlled trial including 162 low-income predominantly Latin-American descendent pregnant women, with less than 32 weeks gestation, and considering breastfeeding. The women were randomly assigned to either a peer-counseling group (PG), or a control group (CG). The women in PG received a community-based exclusive breastfeeding counseling as a breastfeeding promotion strategy, with three prenatal home visits, daily perinatal visits, nine postpartum home visits, and telephone counseling as needed, as well as lactation and education support from the hospital, while women in the CG only received the latter. Descriptive and inferential statistic test were used.
As advised to do in any randomized controlled study, the researchers report absolute and relative frequencies as descriptive statistics for each group, in socio-demographic variables and biomedical factors such as age, marital status, ethnicity, level of education, employment status, parity, etc. They also report mean and standard deviation for birth weight, birth length, and onset of lactation. Assuming that the null hypothesis is true, in other words that there were no differences between the socio-demographic characteristics of the two groups, a chi-squared test or ANOVA was performed. In randomized controlled trials, there should be no statistically significant differences between groups at baseline. It is the way the researcher can be sure that when comparing the groups, the results would not be attributable to these potential differences. Chi-square test is an excellent statistic to compare observed frequencies to expected frequencies, and deciding if the differences are due to chance between two groups. ANOVA, on the other hand, is used to test the significance of group differences between two or more groups, when the independent variable or factor has two or more categories. As opposed to ANOVA, a Student’s t-test is similar and easier to conduct. Nevertheless, the Student’s t-test is recommended to use only when the factor has two levels. Otherwise, the probability of getting a Type I error increases when performing multiple hypothesis test analysis. Therefore, the researchers made a good choice of ANOVA over a Student’s t-test for variables such as gestational age at delivery, infant birth weight, and infant birth length. The rest of the socio-demographic and biomedical variables were analyzed using chi-squared test.
The researchers state, “with the exception of preferred language of interview, pregnancy intentions, planned breastfeeding duration, and infant birth length, there were no significant between-group differences in sociodemographic characteristics at baseline” (Anderson et al., 2005), and then they refer the reader to the first table. However, there is a lack of the actual p value, or confident intervals. The researchers do not even state a significance level. Therefore, as a reader, it is hard to properly evaluate the between-group differences.
Overall, the researchers used very simple but powerful statistic tools to perform their analysis and deliver their results. As the authors state in the discussion section, “findings from this study indicate that the use of trained community-based peer counselors, within the context of a Baby Friendly Hospital, is a very efficacious approach to promote exclusive breastfeeding in the United States” (Anderson et al., 2005). It is very important as a researcher to select the appropriate test that will lead the study to conclusions like this one. Nevertheless, this study might be biased because it was not double blind, and the interviewer knew the study hypothesis. Including the chi-squared test p value, and using a double-blind design are recommendations for future studies following the same methodology.
References
Anderson, A., Damio, G., Young, S., Chapman, D., Pérez-Escamilla, R. (2005). A Randomized Trial Assessing the Efficacy of Peer Counseling on Exclusive Breastfeeding in a Predominantly Latina Low-Income Community. Archives of Pediatric & Adolescent Medicine. 159. 836-841. doi: 10.1001/archpedi.159.9.836
Sistrom, C.L., Garvan, C.W. (2004). Proportions, Odds, and Risk. Radiology. 230. 12-19. doi: 10.1148/radiol.2301031028
