1 Q & A

What are the problems with the previous assignment? Anything not very clear?

2 Projects in Rstudio

Rstudio -> File -> New project …

Projects save the workspace and set working directory. You can work with several R scripts at the same time.

3 👉 Problem set

Using data ESS8PT.sav, answer the following questions:

1. What is the mean satisfaction of life (variable stflife) in Portugal? Is it significantly different from the middle of the scale, 5?
2. Does life satisfaction significantly correlate with subjective health (variable health)?
3. Does life satisfaction significantly differ across genders (variable gndr)?
4. Does life satisfaction significantly differ across types of settlement (domicil)?
5. What are independent effects of gender, age (agea), and type of settlement (domicil)? Build a regression model; check if all the variables have appropriate class (“numeric”, “factor”).
   1. Is there multicollinearity or heteroscedasticity?
   2. What is a predictive power of this model?
   3. What is the impact of every year of age into life satisfaction?
   4. Is there a quadratic (U-curve) effect of age?
   5. Does effect of age differs across genders? (add an interaction term).
   6. Summarize these three models (initial, with squared age, and with interaction) in a single table. Did other coefficients or their significance changed.

Try to write a single R script that works smoothly from the beginning to the end.

4 ⭕️ Bonus: graphical presentation of regressions

library(effects)

eff <- effect(term = c("age*gender"), # the term to plot
              mod  = m1,              # fitted model
              se   = TRUE             # compute standard errors?
              )


plot(x = eff,                      # computed predicted  values
     multiline = TRUE,             # put several lines at the single plot?
     confint=list(style="bands"),  # ribbons (bands) or "lines", or "bars"?
     rug = FALSE                   # plot density by rug bars?
     )

library("sjPlot")

# Just coefficients with confidence intervals
plot_model(m1, type = "est") 

# Interaction terms' slopes
plot_model(m1, type = "int")

# Regression diagnostic plots
plot_model(m1, type = "diag") 

## [[1]]

## 
## [[2]]

## 
## [[3]]

## 
## [[4]]