This seems to me to be a particular feature of psychometrics: it loves rules of thumb that map from the values of a psychometric statistic observed in a dataset to words like “adequate”, “good” or “excellent”. I have increasing clarity that these mappings are deeply problematical and feed into treating questionnaire measures as if they were physical science measures.<\/p>\n\n\n\n
One commonly seen mapping is for Cronbach’s alpha index of internal reliability. As one source has it:
“The most commonly accepted threshold is 0.70.<\/strong> Most dissertation committees will want to see \u03b1 \u2265 0.70 for each scale or subscale you use. Some fields accept 0.60 for exploratory research.<\/em>” (https:\/\/statisticsforresearch.com\/blog\/cronbachs-alpha-explained\/<\/a>). That source also has a very commonly used mapping:<\/p>\n\n\n\n Cronbach's alpha<\/h3>\n
\n\n
\n \n Alpha values <\/td>\n \n Translation <\/td>\n <\/tr>\n \n \n >= .9 <\/td>\n \n Excellent <\/td>\n <\/tr>\n \n \n .8 <= alpha < .9 <\/td>\n \n Good <\/td>\n <\/tr>\n \n \n .7 <= alpha < .8 <\/td>\n \n Acceptable <\/td>\n <\/tr>\n \n \n .6 <= alpha < .7 <\/td>\n \n Questionable <\/td>\n <\/tr>\n \n \n < .6 <\/td>\n \n Poor <\/td>\n <\/tr>\n <\/table>\n<\/div>