Selecting Valid Statistical Test for Evidence Based Medicine Chapter 1 Overview: 1.1 Why Selecting Valid Statistical Tests are Important? 1.2 Factors to be considered for test selection 1.3 Tutorials for selecting valid statistical tests

Cognitive function score at 3 months after ICU discharge Bio-marker (S100) A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A A Pearson’s Correlation P=0.10 (NS ） Spearma’s Correlation P=0.03 （ Significant ） Example 1 ： Different tests → Different results

NIH Research Funds ($ billions) AIDS Breast cancer Ischemic heart disease (ISD) Disability-Adjusted Life-Years (millions) lost due to illness Pearson’s Correlation P=0. ５３９ (NS ） Spearman’s Correlation P ＜０．０００１ （ Significant ） Gross et al. (1999) Example 2 ： Different tests → Different results

Student’s T-test P=0.405 (NS ） Mann-Whitney U test / Wilcoxon rank sum test P=0.012 （ Significant ） APO-E4 Example 3 ： Different tests → Different results

Douglas G. Altman. British Medical Journal, 1994 。 The Scandal of Poor Medical Research What should we think about a doctor who uses the wrong treatment, either willfully or through ignorance, or who uses the right treatment wrongly (such as by giving the wrong dose of a drug)? Most people would agree that such behavior was unprofessional, arguably unethical, and certainly unacceptable. What then would we think about researchers who use the wrong techniques (either willfully or in ignorance), use the right techniques wrongly, misinterpret their results, report their results selectively, cite the literature selectively, and draw unjustified conclusions? We should be appalled. Yet numerous studies of the medical literature, in both general and specialist journals, have shown that all of the above phenomena are common. This is surely a scandal.

Understanding A Statistician’s Mind 1.2 Factors to be considered for test selection

Which type of test do you need: Univaraite or Multivaraite? Question 1 -Are there confounders? - Need Adjustment? RCT vs Observational?

Non Smoker Smoker Confounder Linear Regression with 95.00% Mean Prediction Interval A A A A A A A A A A A A A A A A A A A A Alcohol Consumption % Lung Cancer Graphical Presentation of Confounder – (1) Ignoring confounder

Stratified by Confounder Non smokers Smokers Confounder Linear Regression with 95% Mean Prediction Interval 1234 Alcohol Consumption % Lung Cancer A A A A A A A A A A A A A A A A A A A A Ignoring smoking status falsely detects the association between lung cancer and alcohol consumption. Graphical Presentation of Confounder – (2)

Finding confounders - example

Question 2 Do you want to test for a difference between groups or for correlation between variables? –Comparing mean of two groups? –Correlation between two values?

Question 3 Were the groups paired or unpaired / (dependent or independent)? Are you measuring more than once from one sample?

Question 4 What is the level of measurement for the dependent (outcome) variable? -Ordinal? Disease score (0: normal, 10: abnormal) -Nominal? Example: Gender, Race, Disease/non-disease -Interval (continuous)? Blood pressure, BMI, Weight

Question 5 Is the dependent (outcome) variable normally distributed? If your histogram forms a bell- shaped curve, assume that it is normal; otherwise, assume that it is non-normal.

Question 6 How many groups are there for the independent (predictor) variable? - 2 levels ? - More than 2?

Question 7 What is the total sample size?