## summary - analysis: need to look into outliers - asq: tizina cascaro ianoao (not arrogant way; her approach - suddenly broght) I'm stan developer and would love to get your feedback on one question I had for five years: is it reseacher's role to choose or justify prior? ## my question Regarding Mark's comment "letting ppl making decision" could we frame one research as a hypthotesis testing machine? it's inputs are statistical model and data and output is boundary of parameterized prior that either supports or not supports the hypothesis. This can allow readers with prior belief within success area accept that paper, and those in fail area doesn't accept. Two problems, this may require parameterized prior and second we don't know where the reviewer's prior lies in. --- ## david krack power law and log normal (odds / (odds + 1)) ## mark firm, ceo, industry of consistency btw methodology and theory approach data in pheonomean, of behaviors that model should consider or not Bayesian based on simulation (robustnesss checks); feeling of being threatened can use hypothesis testing - make case (instead of imposing); move away ppl need to understand posterior (probability times my data) - updated conclusion about state of things about data amazon, clinical trials (result for stage 1, .. , ); amazon align theory with method + consistency - we never drew a random sample fortune 500 (but we assume that) - those results are that set of data (need to make argument on generalizable) - most are in the graph (table feel more objective for reviewers) - so much variation in the firms (not all firms are same): buy units and sell units; intercept, sellunits, buyunits, frestruct, restruct, nproduct, nmarket, personnel, alliances, layoffs, hiring - RBV is theory of outlier / - much more accepted in marketing, not in management and econ - readers expect statistical significance test, null hypotheses, simple finding (significance or not) - use everyday examples (bread and butter); failed why don't we use : null hypothesis pay no attention to prior literature (ob developer - job satisfaction; 3k) - prior gives you choices (different ways to formulate priors, ) nht: ronald fisher: murial - random sample of data (sample is not the whole population) -