Takeaways  from the paper : Freedman(1994)

  • Classical Subjectivists will make probability statements about parameters. Even before the data has been collected, the classical subjectivists have an idea about the prior distribution.
  • Radical Subjectivists deny the existence of unknown parameters. For these statisticians, probabilities express degrees of disbelief about observables.
  • Problem with the MLR is that increment in each independent variable is same across all other independent variables . For example an extra year of education is assumed to be worth the same across all levels of experience , both for men and women. This is total crap. BTW, the errors being IID is another big issue
  • Causal inference is a big problem. Instead of fitting model to data, it is better to test the model on a variety of datasets.

image