We used hierarchical linear modeling to account for the nested structure of the data (17). To examine the factors that affect student course grades, we tested a three-level model in which students (level 1) were nested in courses (level 2) and courses were nested within faculty (level 3). The model included partially crossed random effects because students could take courses from more than one faculty member (19). In the model, we controlled for all available student characteristics (gender, race/ethnicity, first-generation status, and SAT scores), all available course characteristics (course enrollment and three dummy variables that account for course level), and all available faculty characteristics (gender, race/ethnicity, age, years of teaching experience, and tenure status). See tables S4 to S6 for correlations among variables at each level. Missing data were handled by listwise deletion. The slope of student race/ethnicity was allowed to vary by course to estimate the cross-level interaction between faculty mindset and student race/ethnicity. The intraclass correlation coefficient (ICC) for course section (level 2) was 0.06, indicating that course sections accounted for 6% of the variance in student grades. The ICC for faculty (level 3) was 0.09, indicating that faculty accounted for 9% of the variance in student grades. The model was fitted using the lme4 package (34) for R version 3.3.1 (35) using restricted maximum likelihood. We used the lmerTest package to obtain P values for fixed effects (36). T tests used the Satterthwaite approximations to degrees of freedom. All continuous variables were standardized. Categorical variables were coded as follows: female = 1, male = 0; URM (Black, Hispanic, Native American) = 1, non-URM (White, Asian) = 0; first-generation = 1, continuing-generation = 0; tenured = 1, nontenured = 0. We added three dummy codes to control for course level, with level 100 as the reference group (i.e., level 200 = 1 and level 100 = 0). Specifically, we estimated a model using the following R code, which was adapted from Bates et al. (34)

M1 <- lmer(Student_Course_Grade ~ Faculty_Mindset*Student_Race

+ Student_Firstgeneration + Student_Gender + Student_SAT

+ Faculty_Gender + Faculty_Teaching_Experience + Faculty_Tenure_Status

+ Faculty_Age + Faculty_Race

+ Course_Enrollment + Course_200Level + Course_300Level + Course_400Level

+ (1 | Student_ID) + (Student_Race |Faculty_ID/Course_ID)

To examine average course evaluations, we tested a two-level model in which courses (level 1) were nested within faculty (level 2). In this model, we controlled for the same course characteristics (course enrollment and three dummy variables that account for class level) and faculty characteristics (gender, race/ethnicity, age, years of teaching experience, and tenure status) as the previous model. The ICC for faculty (level 2) ranged from 0.51 to 0.60, depending on the question, indicating that faculty accounted for approximately 51 to 60% of the variance in students’ course evaluation responses. The following R code was used to estimate the models:

M2 <- lmer(Course_Evaluations ~ Faculty_Mindset

+ Faculty_Gender + Faculty_Teaching_Experience + Faculty_Tenure_Status

+ Faculty_Age + Faculty_Race

+ Course_Enrollment + Course_200Level + Course_300Level + Course_400Level

+ (1|Faculty_ID)

Note: The content above has been extracted from a research article, so it may not display correctly.



Q&A
Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.



We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.