Akademik Veri Yönetim Sistemi
ASEAD INTERNATIONAL SYMPOSIUM SOCIAL SCIENCES, Antalya, Türkiye, 10 - 12 Nisan 2021, ss.125-126
In this study, factors affecting the number of Covid-19 PCR (polymerase chain reaction) tests that
university students have had since March 11, 2020, when the pandemic was declared by World Health
Organization (WHO) were investigated by the method of count data regression analysis. Since having a PCR
test in a fixed time period is a rare event, the number of tests do not normally distributed. Unlike OLS
regression, count data regression does not assume normally distributed residuals with constant variance. In
such cases if the dependent variable is a count variable, count data regression should be used instead OLS
regression. In this study negative binomial regression was used because of the overdispersion problem. The
data were collected via google survey between 9-16 March 2021. All of the respondents are studying at
İstanbul University. The predictors of the count data model are; gender, school grade, age and the family
income. The number of Covid-19 PCR tests taken by the students is the count data variable (dependent
variable). In order to exclude the number of tests that are routinely required by the workplaces, only the
students who do not have a job participated in the survey. As the result of the study, only school grade and age
were statistically significant, family income and gender were not significant. In other words, succesfull and
relatively older students (within the scope of data) have had fewer Covid-19 PCR tests than low success and
relatively younger students.
Keywords: Count Data Regression, Poisson Regression, Overdispersion, Underdispersion,
Negative Binomial Regression.