getLevels {hopit}

Summarize the adjusted and the original self-rated response levels

Description

Summarize the adjusted and the original self-rated response levels.

Usage

getLevels(
  model,
  formula = model$thresh.formula,
  data = model$frame,
  sep = "_",
  decreasing.levels = model$decreasing.levels,
  sort.flag = FALSE,
  weight.original = TRUE
)

Arguments

model

a fitted hopit model.

formula

a formula containing the grouping variables. It is by default set to threshold formula.

data

data used to fit the model.

sep

a separator for the level names.

decreasing.levels

a logical indicating whether self-reported health classes are ordered in increasing order.

sort.flag

a logical indicating whether to sort the levels.

weight.original

a logical indicating if use survey weights for calcualtion of original responses.

Value

a list with the following components:

original

frequencies of original response levels for selected groups/categories.

adjusted

frequencies of adjusted response levels (Jurges 2007 method) for selected groups/categories.

N.original

the number of original response levels for selected groups/categories.

N.adjusted

the number of adjusted response levels (Jurges 2007 method) for selected groups/categories.

categories

selected groups/categories used in summary.

tab

an original vs. an adjusted contingency table.

mat

a matrix with columns: grouping variables, original response levels, adjusted response levels. Each row corresponds to a single individual from the data used to fit the model.

Author(s)

Maciej J. Danko

References

Jurges H (2007). “True health vs response styles: exploring cross-country differences in self-reported health.” Health Economics, 16(2), 163-178. doi:10.1002/hec.1134.

Oksuzyan A, Danko MJ, Caputo J, Jasilionis D, Shkolnikov VM (2019). “Is the story about sensitive women and stoical men true? Gender differences in health after adjustment for reporting behavior.” Social Science & Medicine, 228, 41-50. doi:10.1016/j.socscimed.2019.03.002.

See Also

getCutPoints, latentIndex, standardiseCoef, hopit.

Examples

# DATA
data(healthsurvey)

# the order of response levels decreases from the best health to
# the worst health; hence the hopit() parameter decreasing.levels
# is set to TRUE
levels(healthsurvey$health)

# fit a model
model1 <- hopit(latent.formula = health ~ hypertension + high_cholesterol +
                heart_attack_or_stroke + poor_mobility + very_poor_grip +
                depression + respiratory_problems +
                IADL_problems + obese + diabetes + other_diseases,
              thresh.formula = ~ sex + ageclass + country,
              decreasing.levels = TRUE,
              control = list(trace = FALSE),
              data = healthsurvey)

# Example 1 ---------------------

# calculate a summary by country
hl <- getLevels(model=model1, formula=~ country, sep=' ')
plot(hl, las=1, mar = c(3,2,1.5,0.5))

# differences between frequencies of original and adjusted health levels
round(100*(hl$original - hl$adjusted),2)

# extract good and bad health levels (combined levels)
Org <- cbind(bad = rowSums(hl$original[,1:2]),
             good = rowSums(hl$original[,4:5]))
Adj <- cbind(bad = rowSums(hl$adjusted[,1:2]),
             good = rowSums(hl$adjusted[,4:5]))
round(100*(Org - Adj),2)

# plot the differences
barplot(t(Org - Adj), beside = TRUE, density = 20, angle = c(-45, 45),
        col = c('pink4', 'green2'),
        ylab = 'Original - adjusted reported health frequencies')
abline(h = 0); box()
legend('top', c('Bad health','Good health'),
       density = 20, angle = c(-45, 45),
       fill = c('pink4', 'green2'), bty = 'n', cex = 1.2)

# in country X, bad health seems to be over-reported while good health
# is under-reported; in country Z, good health is highly over-reported.

# Example 2 ---------------------

# summary by gender and age
hl <- getLevels(model = model1, formula=~ sex + ageclass, sep=' ')
plot(hl)

# differences between frequencies of original and adjusted health levels
round(100*(hl$original - hl$adjusted),2)

# extract good health levels (combined "Very good" and "Excellent" levels)
Org <- rowSums(hl$original[,4:5])
Adj <- rowSums(hl$adjusted[,4:5])
round(100*(Org - Adj),2)

pmar <- par('mar'); par(mar = c(9.5, pmar[2:4]))
barplot(Org-Adj,
        ylab = 'Original - adjusted reported good health frequencies',
        las = 3,
        density = 20, angle = c(45, -45), col = c('blue', 'orange'))
abline(h = 0); box(); par(mar = pmar)
legend('top', c('Man','Woman'), density = 20, angle = c(-45, 45),
       fill = c('blue', 'orange'), bty = 'n', cex = 1.2)

# results show that women in general tend to over-report good health,
# while men aged 50-59 greatly under-report good health.

# more examples can be found in the description of the boot_hopit() function.