8.16 Summary of important R code

There is very little “new” R code in this chapter since all these methods were either used in the ANOVA or SLR chapters. The models are more complicated but are built off of methods from previous chapters. In this code, y is a response variable, x1, x2, …, xK are quantitative explanatory variables, group is a factor variable and the data are in DATASETNAME.

  • scatterplot(y~x1|group, data=DATASETNAME, smooth=F)

    • Requires the car package.

    • Provides a scatterplot with a regression line for each group.

  • MODELNAME <- lm(y~ x1+x2+…+xK, data=DATASETNAME)

    • Estimates an MLR model using least squares with \(K\) quantitative predictors.

  • **MODELNAME <- lm(y~ x1*group, data=DATASETNAME)**

    • Estimates an interaction model between a quantitative and categorical variable, providing different slopes and intercepts for each group.

  • MODELNAME <- lm(y~ x1+group, data=DATASETNAME)

    • Estimates an additive model with a quantitative and categorical variable, providing different intercepts for each group.

  • summary(MODELNAME)

    • Provides parameter estimates, overall \(F\)-test, R2, and adjusted R2.

  • par(mfrow=c(2, 2)); plot(MODELNAME)

    • Provides four regression diagnostic plots in one plot.

  • confint(MODELNAME, level=0.95)

    • Provides 95% confidence intervals for the regression model coefficients.

    • Change level if you want other confidence levels.

  • plot(allEffects(MODELNAME))

    • Requires the effects package.

    • Provides a plot of the estimated regression lines with 95% confidence interval for the mean.

  • vif(MODELNAME)

    • Requires the car package.

    • Provides VIFs for an MLR model. Only use in additive models - not meaningful for models with interactions present.

  • predict(MODELNAME, se.fit=T)

    • Provides fitted values for all observed \(x\text{'s}\) with SEs for the mean.

  • predict(MODELNAME, newdata=tibble(x1 = X1_NEW, x2 = X2_NEW, …, xK = XK_NEW, interval=“confidence”)

    • Provides fitted value for specific values of the quantitative predictors with CI for the mean.

  • predict(MODELNAME, newdata=tibble(x1 = X1_NEW, x2 = X2_NEW, …, xK = XK_NEW, interval=“prediction”)

    • Provides fitted value for specific values of the quantitative predictors with PI for a new observation.

  • Anova(MODELNAME)

    • Requires the car package.

    • Use to generate ANOVA tables and \(F\)-tests useful when categorical variables are included in either the additive or interaction models.

  • AIC(MODELNAME_1, MODELNAME_2)

    • Use to get AIC results for two candidate models called MODELNAME_1 and MODELNAME_2.

  • options(na.action = “na.fail”)
    dredge(FULL_MODELNAME, rank=“AIC”)

    • Requires the MuMIn package.

    • Provides AIC and delta AIC results for all possible simpler models given a full model called FULL_MODELNAME.