Interactive graphics Understanding OLS regression Normal approximation to the Binomial distribution
General Stats Software example: OLS regression OLS regressionOLS regression example: Poisson regression Poisson regressionPoisson regression as well as specialized software
Specialized software Testing: Classical test theoryClassical test theory – ITEMIN Item response theoryItem response theory –BILOG-MG –PARSCALE –MULTILOG –TESTFACT
Specialized software Structural equation modeling (SEM) – – –
Specialized software Hierarchical linear modeling (HLM) – –
Open data
Run simple linear regression
Analyze Regression Linear
Enter the DV and IV
Check for confidence intervals
Age accounts for about 37.9% of the variability in Gesell score The regression model is significant, F(1,19) = , p =.002 The regression equation: Y’= X Age is a significant predictor, t(9)=-3.633, p=.002. As age in months at first word increases by 1 month, the Gesell score is estimated to decrease by about points (95% CI: , -.478) Output
Enter the data Fit a Poisson loglinear model: log(Y/pop) = + 1 (Fredericia) + 2 (Horsens) + 3 (Kolding) + 4 (Age) Click to execute
City doesn’t seem to be a significant predictor, whereas Age does. G 2 = 46.45, df = 19, p <.01
Plot of the observed vs. fitted values-- obviously model not fit
Fit another Poisson model: log(Y/pop) = + 1 (Fredericia) + 2 (Horsens) + 3 (Kolding) + 4 (Age) + 5 (Age) 2 Both (Age) and (Age) 2 are significant predictors.
Plot of the observed vs. fitted values: model fits better
Fit a third Poisson model (simpler): log(Y/pop) = + 1 (Fredericia) + 2 (Age) + 3 (Age) 2 All three predictors are significant.
Plot of the observed vs. fitted values: much simpler model
Item Response Theory Easy item Easy item Hard item Hard item Person Ability Item Difficulty Low ability person: easy item - 50% chance
Low ability person: moderately difficult item - 10% chance Item Response Theory Easy item Easy item Hard item Hard item Person Ability Item Difficulty High ability person, moderately difficult item 90% chance
% - 50% - 0% - Probability of success Item Item Response Theory Item difficulty/ Person ability