Presentation on theme: "Data: Quantitative (Histogram, Stem & Leaf, Boxplots) versus Categorical (Bar or Pie Chart) Boxplots: 5 Number Summary, IQR, Outliers???, Comparisons."— Presentation transcript:
Data: Quantitative (Histogram, Stem & Leaf, Boxplots) versus Categorical (Bar or Pie Chart) Boxplots: 5 Number Summary, IQR, Outliers???, Comparisons Center, Shape & Spread! For Skewed Data use…median & IQR For Symmetric data use…mean & SD Normal Model (68 - 95 - 99.7), SD & z- scores
Linear Regression: Explanatory and response variables. Association versus Correlation Correlation Coefficient - strength and direction Residuals R-squared Influential Points (leverage) & outliers Extrapolation Ladder of Powers
Watch out for bias!!! –Voluntary –Convenience –Under-coverage –Non-response –Response
Observational Studies –Retrospective versus prospective Experiments –Random assignment of subjects –Level, factor & treatments –4 principles of experimental design Control extraneous variation Randomize subjects Replicate Block & blind –Statistically significant??? –Control versus placebo –Best experiments are randomized, comparative (control), double-blind & placebo-controlled!!!
–Matched Pairs Match subjects that are similar in ways not under study (ex: right & left feet) –Randomized Block Design –Causation? If only possible explanation for the difference is… –Lurking versus confounding variables
Random Variables –Expected Value –Standard Deviation –Means & Variances Factors –Change mean & variance Addends –Change mean but not variance Combining Random Variables –Means Add or Subtract –Always add variances
Probability Models –Geometric When do we see the first success??? Conditions Expected Value & SD –Binomial k successes in n trials Conditions Expected Value & Standard Deviation
M.C. Tips Underline key words Write out your procedure Dont Rush Balance your time (2-3 min per ???)
F.R. Tips Do easier ???s first Answer the ??? Asked (underline) If miss 1 part, answer a part you know Inference: Conditions/Mechanics/Conclusion Be concise Verify conditions (not just a checklist) 2 decimal places for z-scores, 4 for p-values Context is King!!!