Hello everyone,
I wanted to share some insights about my research on caste bias in college scores among students in India.
In India, the caste system persists, and individuals from lower castes often face discrimination due to their lower status in the societal hierarchy. My research focuses on studying caste bias in the exam scores of college students.
I've been analyzing exam score data from 700 students at an institute in New Delhi. These students are required to complete 24 courses to graduate, each involving two written tests. One test is blindly evaluated by random external evaluators who are not affiliated with the institute, while the other is evaluated by the course instructor.
To measure exam score bias, I'm subtracting the blind score from the non-blind score for each course of each student, resulting in 16,800 observations (700 students x 24 pairs of exams).
Initially, I employed quantile regression to estimate the coefficient of student caste at different quantiles of the subtracted score. However, I encountered an issue where too many observations of the difference in scores were tied.
Could you please advise if quantile regression is still appropriate in this case? My main interest lies in understanding the trend of bias at different quantiles of the difference in scores.
Looking forward to your insights.
Will bootstrapped quantile regression be right approach?
I wanted to share some insights about my research on caste bias in college scores among students in India.
In India, the caste system persists, and individuals from lower castes often face discrimination due to their lower status in the societal hierarchy. My research focuses on studying caste bias in the exam scores of college students.
I've been analyzing exam score data from 700 students at an institute in New Delhi. These students are required to complete 24 courses to graduate, each involving two written tests. One test is blindly evaluated by random external evaluators who are not affiliated with the institute, while the other is evaluated by the course instructor.
To measure exam score bias, I'm subtracting the blind score from the non-blind score for each course of each student, resulting in 16,800 observations (700 students x 24 pairs of exams).
Initially, I employed quantile regression to estimate the coefficient of student caste at different quantiles of the subtracted score. However, I encountered an issue where too many observations of the difference in scores were tied.
Could you please advise if quantile regression is still appropriate in this case? My main interest lies in understanding the trend of bias at different quantiles of the difference in scores.
Looking forward to your insights.
Will bootstrapped quantile regression be right approach?
