1.2 study designs, data types, choosing statistical tests, data analysis checklist
- 2. 1.2 Study Designs, Data Types, Choosing Statistical Tests, Data Analysis Checklist Dr. Mohamed Ayoub Mortagy, MD www.simpleresearch.net info@simpleresearch.net
- 3. Pyramid of evidence Source: wikipedia
- 5. Observational vs interventional (experimental) studies Similarity: both study relation between risk factor / exposure and outcome Observational studies: exposure is not assigned by investigators Interventional studies: exposure is assigned by investigators
- 6. Case Report and case series Discusses only 1 patient case with a rare finding or outcome or atypical case Low evidence No generalization / inference No statistics involved Case report example: https://pubmed.ncbi.nlm.nih.gov/32923205/ Case series example: https://www.cdc.gov/mmwr/preview/mmwrhtml/june_5.htm
- 7. Analytical observational studies No intervention Stronger evidence level than case reports and series Analysis using statistics Measures relation between exposure / risk factor and outcome
- 8. Analytical observational studies Cross-sectional / survey studies (one point of time) Case control study (retrospective) Cohort (prospective or retrospective)
- 10. Cross sectional study design • https://deakin.libguides.com/q uantitative-study-designs/cross- sectional
- 12. Cross sectional Pros and Cons Pros Cheap Simple Quick Ethically safe Cons • Establish association not causality • Recall bias (e.g., surveys) • Confounders • Unequal group sizes
- 15. How to avoid confounding During study design: randomization and/or matching During study analysis: multivariate regression analysis to statistically adjust for confounders
- 16. Case control study Design • https://deakin.libguides.com/ quantitative-study- designs/casecontrol
- 18. Case Control Pros and Cons Pros Quick Cheap Suitable for rare disorders Fewer subjects needed than cross-sectional studies Cons • Rely on recall or record • Confounders • Selection of control group is difficult • Selection bias may be present
- 20. Prospective cohort study design • https://deakin.libguide s.com/quantitative- study- designs/cohortstudies
- 23. Cohort Pros and Cons Pros Ethically safe Matching of subjects Standardized eligibility criteria and outcome assessment Cons • Exposure may be linked to a hidden confounder • Blinding is difficult • No randomization • Not suitable for rare diseases
- 25. Analytical observational studies summary • https://www.nature.com/arti cles/6400436
- 30. RCT Pros and Cons Pros • Blinding • Randomization • less confounding Cons Expensive Time consuming Ethical issues
- 31. Randomized trial design types Parallel trials Cross-over trial Matched pair trial Withdrawal trials Factorial trials
- 36. Randomized controlled trial example https://pubmed.ncbi.nlm.nih.gov/32492084/
- 41. Reviews example Narrative / literature review: https://pubmed.ncbi.nlm.nih.gov/32850602/ Systematic review: https://pubmed.ncbi.nlm.nih.gov/32725955/ Metanalysis: https://pubmed.ncbi.nlm.nih.gov/32860962/
- 46. Quantitative (numerical data) Types Continuous Discrete (counts) Examples: Age, blood pressure, BMI, Pulse
- 47. Categorical Binary examples: dead / alive – treatment / placebo – disease / no disease – exposed / not exposed Nominal (unordered) examples: blood types – marital status Ordinal (ordered) examples: cancer staging – ratings on likert scale – age in categories
- 48. Time to event variables Time taken for an event to happen Examples: time to death – time to heart attack – time to chronic kidney disease – time to cancer mortality
- 50. Dependent vs. independent variables • https://www.tobiipro.co m/learn-and- support/learn/steps-in-an- eye-tracking- study/design/what-are- experimental-variables/
- 51. Choosing statistical test Outcome variable Independent or Paired observations? Non parametric tests Independent Paired Continuous T test (2 groups) ANOVA (> 2 groups) Linear correlation Linear regression Paired T test Repeated measures anova GEE model Wilcoxon signed rank test Wilcoxon rank sum test Kruskal wallis test Spearman rank correlation coefficient Binary or categorical Risk difference or relative risk Chi square test Logistic regression Mcnemar’s test Conditional logistic regression Fischer’s exact test Mcnemar’s exact test Counts Rate ratio Poisson regression GEE model Negative binomial regression Time to event Rate ratio Kaplan Meier statistics Cox regression Counting process model Time varying effects
- 53. Data analysis check list Prepare the data 1 Check the data 2 Study your variables 3 Deal with missing data 4 Explore simple relations and assumptions 5 Prepare a descriptive table 6 Test the main hypothesis 7 Build final models and accompanying graphics 8 Perform sensitivity analysis 9 Prepare final code and procedure documents 10