Antisemitism: Change in antisemitism after seeing a movie

Description

A classical data set that has been used several times in the past, but not analyzed by means of the methods advocated in this book (Glock, 1955; Campbell & Clayton, 1961; Hagenaars, 1990, pp. 215-233, and Hagenaars, 1990, Section 5.3). The data are from a panel study among 503 white Christians living in and around Baltimore. The study's purpose was to determine the effect of seeing the film `Gentleman's Agreement' on reducing the level of antisemitism (Glock, 1955, p. 243). Antisemitism was measured in November 1947 (variable $A$ ) prior to the movie being locally shown and consisted of three categories : 1 = high, 2 = moderate, and 3 = low. Antisemitism was measured again in May 1948 (variable $B$ ). In addition, the respondents were asked whether or not they had (voluntary) seen the movie, which had been shown in Baltimore theaters during the period between the two interviews (variable $X$ ). The experimental group (with $X = 1$ ) consisted of those respondents who saw the movie; the control group (with $X = 2$ ) consisted of those who did not. The data are tabulated in Bergsma, Croon, and Hagenaars (2009, Table 5.8).

Section 5.2.2 in Bergsma, Croon, and Hagenaars (2009).

Usage

data(GSS93)

Arguments

Format

A data frame with 496 observations on the following three variables.

X: Seen the film (factor): 1 = Seen; 2 = Not seen;
A: Antisemitism at Time 1 (ordered): 1 = High; 2 = Moderate; 3 = Low.
B: Antisemitism at Time 2 (ordered): 1 = High; 2 = Moderate; 3 = Low.

References

Bergsma, W. P., Croon, M. A., & Hagenaars, J. A. P. (2009). Marginal models for dependent, clustered, and longitudunal categorical data. Berlin: Springer.

Campell & Clayton (1961)

Glock (1955)

Hagenaars, 1990

Examples

Run this code

data(Antisemitism)

## Sample marginal distributions
# at applied to data gives vectorized 2x2x3 table TXR (Time x Seen film or not x Response)
at <- MarginalMatrix(c("X","A","B"), list(c("X","A"), c("X","B")), c(2,3,3));
stats = SampleStatistics(
  dat = Antisemitism, 
  coeff = at, 
  Labels = c("T","X","R"), 
  CoefficientDimensions = c(2,2,3))

## Models for table XR given T

# at1 applied to data gives vectorized conditional 2x3 table XR (XR conditional on T<-1)
at1 <- MarginalMatrix(c("X", "A", "B"), list(c("X", "A")), c(2, 3, 3));
# at2 applied to data gives vectorized conditional 2x3 table XR (XR conditional on T<-2)
at2 <- MarginalMatrix(c("X", "A", "B"), list(c("X", "B")), c(2, 3, 3));

bt1 <- ConstraintMatrix(c("X", "R"), list(c("X"), c("R")), c(2, 3));
bt2 <- ConstraintMatrix(c("X", "R"), list(c("X"), c("R")), c(2, 3));

model1 <- list(bt1, "log", at1);
model2 <- list(bt2, "log", at2);

# model1 doesn't converge, I don't know the reason and am trying to find out 
# (it does converge in the Mathematica programme).
fit = MarginalModelFit(
 dat = Antisemitism, 
 model = model2, 
 Labels = c("X","R"), 
 CoefficientDimensions = c(2,3), 
 MaxSteps=100, 
 ShowProgress=10,
 MaxStepSize=.5)