2019 acs graduate Student Survey Report

Analyses of Subgroup Differences
In addition to looking at the responses from all students, differences among groups were calcu-
lated and tested for statistical significance. Specifically, they focused on differences by degree 
(master’s and doctoral) and differences by gender (male and female respondents). We also ex-
amined differences by gender in the student–advisor pair for questions involving student–advisor 
relationships and advisors’ behaviors. Additional comparative analyses, including examinations 
by year in program, age, citizenship status, and underrepresented racial/ethnic minority (URM) 
status, were conducted as deemed relevant.
In order to determine whether differences among groups were statistically significant, a number 
of statistical tests were used. Depending on the survey item, these included independent sample 
t-tests, z-tests for proportions, and/or Pearson’s chi-squared test for independence. For these
tests, a p-value of less than .05 denoted statistical significance (the size of a given difference

---

44
also was considered to guide understanding of variation—however, effect sizes were not formal-
ly computed for the purpose of this report). Chi-squared tests were used primarily to determine 
whether responses across ordinal or nominal response categories were distributed differently 
among groups. To perform pairwise comparison among groups of two or more, pairwise Chi-
squared tests were used. Z-tests were used to determine whether the proportion of respons-
es for an item that had more than two response options differed among groups. Finally, when 
comparing means among groups, independent sample t-tests were used. Statistical tests were 
robust in the face of differing sample sizes. In the body of the report, all noted differences among 
groups are significant at p<.05.

---

45