2016 AAPM Annual Meeting
Back to session list

Session Title: Statistical Failings That Keep Us All In The Dark
 Question 1: Various publications have found a higher than expected incidence of p-values immediately below p=0.05 as compared to p-values immediately above p=0.05. Some possible reasons for this over-representation could include: Reference: B. Ginsel, et al., The distribution of probability values in medical abstracts: an observational study., BMC Res Notes, Nov 26, 2015. Choice A: Publication bias. Choice B: Statistical fraud. Choice C: Methodological errors (selective reporting, selective analyses, underpowered analysis). Choice D: A and B. Choice E: All of the above. Question 2: Which is the most appropriate definition of a p-value? Reference: R.L. Wasserstein and N.A. Lazar, The ASA’s statement on p-values, context, process, and purpose, The American Statistician, accepted version published online 3/7/2016. Choice A: A p-value is the probability that the null hypothesis is true Choice B: A p-value is the probability under a specified statistical model that a statistic al summary of the would be equal to or more extreme than its observed value. Choice C: A p-value is the probability that the result was a random coincidence Choice D: A p-value is the probability the results will not hold up if the experiment is repeated Choice E: All of the above. Question 3: Effect sizes are more important than p-values because: Reference: Source: http://amstat.tandfonline.com/doi/abs/10.1080/00031305.2016.1154108 Choice A: They are unitless measures. Choice B: They are sample independent. Choice C: They tell you how practically meaningful your results are. Choice D: They are unrelated to the p-value. Choice E: C and D. Question 4: In non-mathematical terms, the problem of performing multiple comparisons without controlling for them is that it: Reference: H. Motulsky, Intuitive Biostatistics: A nonmathematical guide to statistical thinking, Oxford Univ Press, (2014), pp. 183. Choice A: Makes it difficult to detect cause from correlation among all of the predictors. Choice B: Makes it less likely to achieve a p-value <= 0.05 among all of the predictors. Choice C: Across all of the predictors, leads to an increased probability of erroneously finding a difference that is statistically significant when in fact all of the null hypotheses are true. Choice D: Makes it more difficult to compile the analysis due to the number of results presented. Choice E: Is difficult to achieve with currently available statistical software. Question 5: The reason censoring tick marks are important for the interpretation of a survival curve is: Reference: J. T. Rich, et al., A Practical Guide to Understanding Kaplan-Meier Curves, Otolaryngol Head Neck Surg 143(3), 2010, 331-336. Choice A: They represent the total fraction of subjects lost to followup at a given timepoint. Choice B: They yield important information about the number of subjects remaining in the study (and thus the reliability of the curve) at each timepoint. Choice C: They represent the timepoint when a subject reaches the specified event being analyzed. Choice D: They represent fixed time intervals on the KM curve. Question 6: The best approach to missing data is: Reference: http://www.valueinhealthjournal.com/article/S1098-3015(10)60635-3/references Choice A: Last value carried forward. Choice B: A sensitivity analysis of competing approaches. Choice C: Simple imputation. Choice D: Multiple imputation. Choice E: Complete Cases.
Back to session list