Zarrukh Rakhimov,
PhD candidate in Econometrics and Statistics
Email: [email protected]
Westminster International University in Tashkent
Istiqbol str. 12, 100047 Tashkent, Uzbekistan
ORCID: 0009-0001-0583-4819
Abstract: Linear regression is one of the widely used statistical methods in social sciences. The core part of the regressions are coefficients, which bring some inference. Yet, we rely on hypothesis testing or confidence intervals and certain assumptions underlying linear models such as sample size being large enough. In this study, we suggest alternative way of constructing confidence intervals using bootstrap, which is expected to work well even when the sample size is smaller than required per OLS assumptions. We find that even in small samples, bootstrap confidence intervals can perform better than traditional interval estimations due to larger interval size
Keywords: sample size, linear model, confidence Interval, bootstrap, accuracy, interval size