Week 11 workbook and solutions

Import the jittered NZAVS dataset

Show code

### Libraries
library("tidyverse")
library("patchwork")
library("lubridate")
library("kableExtra")
library("gtsummary")
library("lubridate")
library("equatiomatic")
library("ggdag")
library("brms")
library("rstan")
library("rstanarm")
library("bayesplot")
library("easystats")
library("kableExtra")
library("broom")
library("tidybayes")
library("bmlm")
# rstan options
rstan_options(auto_write = TRUE)
options(mc.cores = parallel::detectCores ())
theme_set(theme_classic())

Show code

# read data

nz_0 <- as.data.frame(readr::read_csv2(
  url(
    "https://raw.githubusercontent.com/go-bayes/psych-447/main/data/nzj.csv"
  )
))

# to re-level kessler 6 variables
f <-
  c(
    "None Of The Time",
    "A Little Of The Time",
    "Some Of The Time",
    "Most Of The Time",
    "All Of The Time"
  )

# get data into shape
nz_cr <- nz_0 %>%
  dplyr::mutate_if(is.character, factor) %>%
  dplyr::select(
    -c(
      SWB.Kessler01,
      SWB.Kessler02,
      SWB.Kessler03,
      SWB.Kessler04,
      SWB.Kessler05,
      SWB.Kessler06
    )
  ) %>% 
  dplyr::mutate(Wave = as.factor(Wave)) %>%
  dplyr::mutate(FeelHopeless = forcats::fct_relevel(FeelHopeless, f)) %>%
  dplyr::mutate(FeelDepressed = forcats::fct_relevel(FeelDepressed, f)) %>%
  dplyr::mutate(FeelRestless = forcats::fct_relevel(FeelRestless, f)) %>%
  dplyr::mutate(EverythingIsEffort = forcats::fct_relevel(EverythingIsEffort, f)) %>%
  dplyr::mutate(FeelWorthless = forcats::fct_relevel(FeelWorthless, f)) %>%
  dplyr::mutate(FeelNervous = forcats::fct_relevel(FeelNervous, f)) %>%
  dplyr::mutate(Wave = as.factor(Wave)) %>%
  dplyr::mutate(male_id = as.factor(Male)) %>%
  dplyr::mutate(date = make_date(year = 2009, month = 6, day = 30) + TSCORE) %>%
  dplyr::mutate(
    FeelWorthless_int = as.integer(FeelWorthless),
    FeelNervous_int =  as.integer(FeelNervous),
    FeelHopeless_int =  as.integer(FeelHopeless),
    EverythingIsEffort_int =  as.integer(EverythingIsEffort),
    FeelRestless_int =  as.integer(FeelRestless),
    FeelDepressed_int =  as.integer(FeelDepressed),
    HLTH.Fatigue_int = as.integer(HLTH.Fatigue + 1)
  ) %>%
  dplyr::mutate( yearS = (TSCORE - min(TSCORE)/365) ) %>%
  dplyr::mutate(KESSLER6sum = as.integer(KESSLER6sum))

## if you do anything with covid (warning, such a model would be tricky)

ord_dates_class <- c("Baseline",
                     "PreCOVID",
                     "JanFeb",
                     "EarlyMarch",
                     "Lockdown",
                     "PostLockdown")

nzl <- nz_cr %>%
  dplyr::filter(YearMeasured == 1) %>%
  dplyr::mutate(yearS = (TSCORE - min(TSCORE)/365)) %>%
  dplyr::mutate(WSCORE = as.factor(WSCORE)) %>%
  dplyr::mutate(Covid_Timeline =
                  as.factor(ifelse(
                    TSCORE %in% 3896:3921,
                    # feb 29 - march 25th
                    "EarlyMarch",
                    ifelse(
                      TSCORE %in% 3922:3954,
                      "Lockdown",
                      #march 26- Mon 27 April 2020
                      ifelse(
                        TSCORE > 3954,
                        # after april 27th 20202
                        "PostLockdown",
                        ifelse(
                          TSCORE %in% 3842:3895,
                          # jan 6 to feb 28
                          "JanFeb",
                          ifelse(TSCORE %in% 3665:3841 &
                                   Wave == 2019,
                                 "PreCOVID",
                                 "Baseline"  # 3672 TSCORE or  20 July 2019))))))))
                          )
                        )
                      )
                    ))))%>%
  dplyr::mutate(Covid_Timeline = forcats::fct_relevel(Covid_Timeline, ord_dates_class))%>%
  dplyr::mutate(Id = factor(Id))

Assessment: briefly answer three of the following five questions.

1. What is the difference between prediction and causal inference: give an example of how regression may be use useful for each task. Make sure your explanation includes a DAG.

2. Explain “collider confounding” (or “collider bias”), and explain how collider bias can spoil inference. Make sure that your explanation includes a DAG.

3. Using the nzl dataset, select at least five demongraphic/ideological variables that might be related to an exposure variable and and outcome variable of your choice. Create a DAG and identify which variables you should include to obtain an unbiased estimate of the causal effect of your exposure variable on your outcome variable. Test your model and interpret the results.

4. Write a DAG for one of your previous workbook regression analyses; Explain your DAG. Redo your analysis to obtain an unbiased causal inference (conditional on your DAG). Compare the results of your current model with the results of your previous model and note any differences in interpretation.

5. Examine the causal assumptions of a psychological theory for which you have data. Test the causal assumptions of this model. Make sure to draw a DAG. Consider simulating data if/where relevant.

Marking criteria

1. Clarity, accuracy, initiative, and organisation in the descriptive component of your work.

2. Clarity, accuracy, initiative, and organisation in your DAG and, if relevant, your regression model(s).

3. Accuracy and insight in the interpretation of DAG, and if relevant, your regression model(s).

Supplement: NZ-jitter information:

Link to the NZAVS data dictionary is here

Link to questions only here