PSYC 3031 Lecture Notes - Lecture 3: Local Regression, Diminishing Returns
Document Summary
Look at data before analyzing, stats methods can only tell you very specific things. Only by looking at it visually, you become very familiar with it + know it, including which statistical techniques that might be appropriate + valuable to use. Using graphs are great esp grouped histogram. So(cid:373)eti(cid:373)es (cid:455)ou do(cid:374)"t wa(cid:374)t s(cid:373)oothi(cid:374)g to (cid:373)ake a poi(cid:374)t. Can look at more than 2 groups with pie charts terrible for data display. Box and whisker: q1 and q4 is 1. 5xiqr. Scatterplots appear simple but can be deceptively complicated, relationship you see may not be generated by the variables you think you are studying (confounds) Beware of making snap judgments about relationship between variables, especially causal ones, on the basis of simple scatterplots, there are invisible variables behind every x and y axis. 2 heterogeneous groups form together to make an apparent relationship. Admission figures example (less % of women admitted overall, but in certain departments there seems to be the opposite)