PSYCH291 Lecture Notes - Lecture 17: Brian Nosek, False Positive Rate, Statistical Significance
16 Dec 2018
School
Department
Course
Professor
Document Summary
Issue in reproducing results flaws getting addressed, on its way to improving. Can inflate false positives (type 1 error) Play with data to achieve statistical significance. Stop collecting data when p <0. 05, even if nothing significant in nature increases false positive timing outliers: No set standards set for two to deal with outliers. Can produce false, statistically significant results flexibility in data analysis/reporting can inflate p-value inflated false positive rate: 61% chance to receive false positive if you are really flexible with data. p-curves: Unlikely you will find p values near p=0. 04 and p =0. 05 mark. too many improbable values of p < 0. 01. Not publishing study when no significant results (journal bias) researchers manipulate data to get their study published in a journal. Brian nosek and colleagues set out to replicate hundreds of published articles to get some results. Published study that showed low replication rate. 47% in biology. external validity issues (social psych)
