Rrepest: Estimation using replicate weights

In Rrepest: An Analyzer of International Large Scale Assessments in Education

Estimation using replicate weights

Description

Estimates statistics using replicate weights (Balanced Repeated Replication (BRR) weights, Jackknife replicate weights,...), thus accounting for complex survey designs in the estimation of sampling variances. It is designed specifically to be used with the data sets produced by the Organization for Economic Cooperation and Development (OECD), some of which include the Programme for the International Assessment of Adult Competencies (PIAAC), Programme for International Student Assessment (PISA) and Teaching and Learning International Survey (TALIS) data sets, but works for any educational large-scale assessment and survey that uses replicated weights. It also allows for analyses with multiply imputed variables (plausible values); where plausible values are used, average estimator across plausible values is reported and the imputation error is added to the variance estimator.

Usage

Rrepest(
  data,
  svy,
  est,
  by = NULL,
  over = NULL,
  test = FALSE,
  user_na = FALSE,
  show_na = FALSE,
  flag = FALSE,
  fast = FALSE,
  tabl = FALSE,
  average = NULL,
  total = NULL,
  coverage = FALSE,
  invert_tests = FALSE,
  save_arg = FALSE,
  cores = NULL,
  ...
)

Arguments

data	(data frame) Data to analyse
svy	(string) Declares the survey settings. It must be equal to one of the following: ALL, IALS, ICCS, ICILS, IELS, PBTS, PIAAC, PIRLS, PISA, PISAOOS, PISA2015, SSES, SSES2023, SVY, TALISSCH, TALISTCH, TALISEC_LEADER, TALISEC_STAFF, TIMSS.
est	(est function) Specifies the estimates of interest. It has three arguments: statistics type, target variable and an (optional) regressor list in case of a linear regression.
by	(string vector) Produces separate estimates by levels of the variable(s) specified by the string vector.
over	(string vector) Requests estimates to be obtained separately for each level of categorical variable(s) identified by the string vector.
test	(bool) if TRUE: Computes the difference between estimates obtained for the lowest and highest values of the 'over' variable(s). (See 'over' option above.) It is useful to test for differences between dependent samples (e.g. female-male).
user_na	(bool) if TRUE: Shows the nature of user defined missing values.
show_na	(bool) if TRUE: Includes missing values (i.e. NAs) when estimating frequencies for the variable of interest.
flag	(bool) if TRUE: Replaces estimation results that are based on fewer observations than required for reporting with NaN. When used with the PIAAC survey settings, it checks if each estimation result is based on at least 30 observations. When used with the PISA, PISAOOS, PISA2015 survey settings, it checks if each estimation result is based on at least 30 observations and 5 schools. When used with the TALISSCH survey settings, it checks if each estimation result is based on at least 10 schools. When used with the TALISTCH survey settings, it checks if each estimation result is based on at least 30 observations and 10 schools.
fast	(bool) if TRUE: Computes estimates by using only 6 replicated weights.
tabl	(bool) if TRUE: Creates customisable and transferable tables using the flextable R package.
average	(grp function) Computes an arithmetic average (or weighted average). It has three arguments: name of the average, column/variable used for computing the average, rows/observations included in the average. It has three arguments: name of the group, column/variable used for computing the group, rows/observations included in the group.
total	(grp function) Computes an average weighted by the estimated size of the target population covered.
coverage	(bool/numeric) TRUE: shows column next to se. Numeric: Shows NaN if bellow the set coverage.
invert_tests	(bool) Invert test columns from Rrepest test = TRUE by name on "b." and "se." in the column name and by sign (*-1) on "b."
save_arg	(bool) TRUE: returns a named list with the estimation data frame and all arguments used in Rrepest.
cores	(numeric) NULL: Will recruit max-1 cores when doing PVs. Else, will recruit the specified number of cores for PVs
...	Other optional parameters include: isced = Filters the data used for analysis by ISCED level (e.g. isced = 2), n.pvs = Customizes the number of plausible values used in the estimation (e.g. n.pvs = 5), cm.weights = Customizes the weights used in the estimation (e.g. cm.weights = c("finw",paste0("repw",1:22))), var.factor = Customizes the variance factor used in the estimation (e.g. var.factor = 1/(0.5^2)), z.score = qnorm(1-0.05/2)

Value

Data frame containing estimation "b." and standard error "se.".

Examples

data(df_pisa18)

Rrepest(data = df_pisa18,
svy = "PISA2015",
est = est("mean","AGE"),
by = c("CNT"))

Rrepest documentation built on April 4, 2025, 2:07 a.m.