Bridging logistic and OLS regression

Kapsalis, Constantine (2010): Bridging logistic and OLS regression.

Abstract

There is broad consensus that logistic regression is superior to ordinary least squares (OLS) regression at predicting the probability of an event. OLS is still widely used in binary choice models because its coefficients are easier to interpret, while the resulting estimates tend to be close to the logit estimates anyway. Although some statistical software provide an easy way of calculating marginal effects (equivalent in interpretation to OLS coefficients) this is not always the case. This paper shows a simple way of calculating marginal effects from logistic coefficients.

• Item Type: MPRA Paper
• Language: English
• Keywords: regression analysis
• Subjects: C - Mathematical and Quantitative Methods > C3 - Multiple or Simultaneous Equation Models ; Multiple Variables > C35 - Discrete Regression and Qualitative Choice Models ; Discrete Regressors ; Proportions

References

Amemiya, T. (1981). "Qualitative response models: a survey", Journal of Economic Literature 19: 1483-1536.

Goldberger, A. (1964). Econometric theory (Wiley, New York).

Moffitt, Robert (1999). “New Developments in Econometric Methods for Labor Market Analysis”, in Handbook of Labour Economics, Volume 3, Chapter 24, Edited by O. Ashenfelter and D. Card.

Pohlmann, John T. and Dennis W. Leitner (2003). “A Comparison of Ordinary Least Squares and Logistic Regression”, The Ohio Journal of Science, 103 (5): 118-125.

Theil, H. (1981). Principles of econometrics (Wiley, New York).