Applies any lsastrat estimator (social_gradient(), segregation_index(),
ieop(), variance_decomposition()) separately within each level of a
grouping variable and stacks the pooled coefficients into one tidy,
comparable table. This is the workhorse for comparative large-scale
assessment research: one call returns the gradient (or any other quantity)
for every country, with confidence intervals, ready to plot or tabulate.
Arguments
- data
A data frame of student-level records.
- by
Name of the grouping column (e.g. a country identifier).
- estimator
The estimator to apply: a function such as
social_gradient, or its name as a string.- ...
Arguments passed on to
estimator(everything exceptdata).- .progress
Logical; message each group as it is processed.
Value
An object of class "lsastrat_by": a list with results (a data
frame with a group column followed by term and the pooled coefficient
columns) plus metadata. Groups for which the estimator fails are dropped
with a warning.
See also
plot.lsastrat_by() for forest/caterpillar plots.
Examples
data(pisa_mini)
# gradient by immigrant background (any grouping column works)
lsa_by(pisa_mini, by = "IMMIG", estimator = social_gradient,
achievement = paste0("PV", 1:10, "MATH"), escs = "ESCS",
weight = "W_FSTUWT", repweights = paste0("W_FSTURWT", 1:64))
#> `social_gradient` by IMMIG (3 groups)
#>
#> group term estimate se statistic df p ci_lo ci_hi
#> native slope 41.867 2.070 20.224 580.355 0.000 37.801 45.933
#> native strength 20.192 1.934 10.438 346.181 0.000 16.388 23.997
#> native curvature 0.752 1.920 0.392 242.884 0.696 -3.031 4.534
#> second_gen slope 39.234 3.548 11.058 792.323 0.000 32.269 46.198
#> second_gen strength 17.931 2.935 6.108 425.972 0.000 12.161 23.700
#> second_gen curvature 0.560 2.792 0.201 1125.479 0.841 -4.919 6.039
#> first_gen slope 32.697 11.252 2.906 338.768 0.004 10.566 54.829
#> first_gen strength 13.126 9.079 1.446 258.647 0.149 -4.752 31.003
#> first_gen curvature 7.279 8.914 0.817 600.676 0.414 -10.227 24.785