ILSAstats (version 0.4.5)

repglm: Generalized Linear Models with Replicate Weights

Description

Fits a generalized linear model using glm for replicate weights. For a detailed explanation on how the standard errors are estimated see repse.

Usage

repglm(
  formula,
  family = stats::gaussian,
  pvs = NULL,
  relatedpvs = TRUE,
  quiet = FALSE,
  summarize = TRUE,
  setup = NULL,
  df,
  wt,
  repwt,
  group = NULL,
  exclude = NULL,
  na.action = getOption("na.action"),
  method
)

Value

a list with the standard errors and the total weights models.

Arguments

formula

an object of class "formula" (or one that can be coerced to that class): a symbolic description of the model to be fitted. The details of model specification are given under ‘Details’.

family

a description of the error distribution and link function to be used in the model. For glm this can be a character string naming a family function, a family function or the result of a call to a family function. For glm.fit only the third option is supported. (See family for details of family functions.)

pvs

if plausible values are not used, this should be NULL. Otherwise it is a list indicating which variables from formula should be replaced by which plausible values variables. For more details check the examples.

relatedpvs

a logical value indicating if pvs are drawn from the same model. If TRUE (default), a total of \(n\) estimations will be done, where \(n\) is the number of plausible values for each plausible value variable. If FALSE, a total of \(n_1 \times n_2 \times n_...\) estimations will be done, where \(n_i\) is the number of plausible values in each plausible value variable.

quiet

a logical value indicating if progress status should be shown while estimating models by group. Default is FALSE.

summarize

a logical value indicating if lm objects should be converted to summary.lm or summary.glm objects and stripped from certain elements to reduce the size of the output object. Default is TRUE.

setup

an optional list produced by repsetup.

df

a data frame.

wt

a string specifying the name of the column (within df) with the total weights.

repwt

a string indicating the common names for the replicate weights columns (within df), or a data frame with the replicate weights.

group

a string specifying the variable name (within df) to be used for grouping. Categories in group are treated as independent, e.g., countries.

exclude

a vector indicating which groups (in the same format as group) should be excluded from the pooled and composite estimates.

na.action

a function which indicates what should happen when the data contain NAs. The default is set by the na.action setting of options, and is na.fail if that is unset. The ‘factory-fresh’ default is na.omit. Another possible value is NULL, no action. Value na.exclude can be useful.

method

a string indicating the name of the replication method. Available options are: "JK2-full", "JK2-half", "FAY-0.5", and "JK2-half-1PV".

Additionally, ILSA names can be used, defaulting into:

• "TIMSS", "PIRLS", or "LANA" for "JK2-full";

• "ICILS", "ICCS", or "CIVED" for "JK2-half";

• "PISA" or "TALIS" for "FAY-0.5";

• and "oldTIMSS", "oldPIRLS", or "RLII" for "JK2-half-1PV".

Note that "oldTIMSS" and "oldPIRLS" refer to the method used for TIMSS and PIRLS before 2015, where within imputation variance is estimated using only 1 plausible value.

Examples

# Less data for shorter example
repdata2 <- repdata[1:200,]

# Creation of replicate weights
RW <- repcreate(df = repdata2, # the data frame with all the information
                wt = "wt", # the total weights column name
                jkzone = "jkzones", # the jkzones column name
                jkrep = "jkrep", # the jkreps column name
                repwtname = "REPWT", # the desired name for the rep weights
                reps = 50, # the number of replications
                method = "ICILS") # the name of the method aka the study name

### No groups ----

# Simple regression - default family = gaussian
repglm(formula = GENDER ~ 1 + Math1,
        family = gaussian, # Link function
        wt = "wt", # Name of total weight column within df
        repwt = RW, # Data frame of weights
        df = repdata2, # Data frame
        method = "ICILS") # the name of the method aka the study name


# Simple regression - change link function
repglm(formula = GENDER ~ 1 + Math1,
        family = quasibinomial, # Link function
        wt = "wt", # Name of total weight column within df
        repwt = RW, # Data frame of weights
        df = repdata2, # Data frame
        method = "ICILS") # the name of the method aka the study name

# Multiple regression
repglm(formula = GENDER ~ 1 + Math1 + Reading1,
        family = quasibinomial, # Link function
        wt = "wt", # Name of total weight column within df
        repwt = RW, # Data frame of weights
        df = repdata2, # Data frame
        method = "ICILS") # the name of the method aka the study name

# Multiple regression - with PVs
## Named list, with element names matching formula variables
pvs = list(Math = paste0("Math",1:3))
pvs

repglm(formula = GENDER ~ 1 + Math + Reading1, # Math1 now is "Math"
        family = quasibinomial, # Link function
        wt = "wt", # Name of total weight column within df
        repwt = RW, # Data frame of weights
        df = repdata2, # Data frame
        pvs = pvs, # Named list
        method = "ICILS") # the name of the method aka the study name