ibn Harm

Tag: Frequentists

  • Frequentists and bayes (3)

    I was revisiting the first articles on this topic and it was not really clear to me what I meant. Thus this entry is sort of a mental note for me.

    As I understand it researchers who prefer frequentists methodologies do not speculate on events which researchers choosing Bayes’ mathematics will state have “probabilities.”

    So before a soccer match is played frequentists can say things like: “based on the previous data, if we assume previous data can say something about future events, we expect team a to win the match.”

    Researchers using Bayes on the other hand might say things like: “based on the previous data we expect the score to become ‘3-1’ (yielding team a as matchwinner).”

    Frequentists should not be able to predict “net results” or “expected goals” I think.

  • Bayes and frequentists #2

    In the earlier post I discussed merely the Bayesian camp so today we will right that wrong.

    Frequentist statistics is helpful in that it enables to arive at general conclusions about a test group in a controlled environment.

    Can frequentist statistics be used to make decisions? It can and is being done for example in computer science viz. a-b testing for example in intelligent interaction design which is a discipline aimed at improving user experience vis-a-vis electronic interfaces.

    In this particular example Bayes would not have sufficed, at least not in its normal form. A applied methodology can be imagined in which Bayesian strength is determined before and after a use-case for two or multiple interfaces.

    The research question would be slightly different from that of the frequentist scientists viz. frequentists have a straightforward method in testing whether there is a significant difference in user experience.

    Since Bayesian inference is designed to arrive at degree of belief with regard to prior probability it is more applicable to individual use cases rather than coming to conclusions for a predetermined test group.

    Hypothesis: frequentist statistics is used to draw general conclusions from carefully selected test groups, bayesian statistics is used to draw case-specific conclusions from historical data and particular context.