Run APC-I model
apci( outcome = "inlfc", age = "acc", period = "pcc", cohort = NULL, weight = NULL, covariate = NULL, data, family = "quasibinomial", dev.test = TRUE, print = TRUE, gee = FALSE, id = NULL, corstr = "exchangeable", unequal_interval = FALSE, age_range = NULL, period_range = NULL, age_interval = NULL, period_interval = NULL, age_group = NULL, period_group = NULL, ... )
outcome: An object of class character containing the name of the outcome variable. The outcome variable can be continuous, categorical, or count.
age: An object of class character representing the age group index taking on a small number of distinct values in the data. Usually, the vector should be converted to a factor (or the terms of "category" and "enumerated type").
period: An object of class character, similar to the argument of age, representing the time period index in the data.
cohort: An optional object of class character representing cohort membership index in the data. Usually, the cohort index can be generated from the age group index and time period index in the data because of the intrinsic relationship among these three time-related indices.
weight: An optional vector of sample weights to be used in the model fitting process. If non-NULL, the weights will be used in the first step to estimate the model. Observations with negative weights will be automatically dropped in modeling.
covariate: An optional vector of characters, representing the name(s) of the user-specified covariate(s) to be used in the model. If the variable(s) are not found in data, there will be an error message reminding the users to check the data again.
data: A data frame containing the outcome variable, age group indicator, period group indicator, and covariates to be used in the model. If the variable(s) are not found in data, there will be an error message reminding the users to check the input data again.
family: Used to specify the statistical distribution of the error term and link function to be used in the model. Usually, it is a character string naming a family function. For example, family can be "binomial", "multinomial"", or "gaussian". Users could also check R package glm for more details of family functions.
dev.test: Logical, specifying if the global F test should be implemented before fitting the APC-I model. If TRUE, apci will first run the global F test and report the test results; otherwise, apci will skip this step and return NULL. The default setting is TRUE. However, users should be aware that the algorithm will not automatically stop even if there is no significant age-by-period interactions based on the global F test.
print: Logical, specifying if the intermediate results should be displayed in the console when fitting the model. The default setting is TRUE to display the results of each procedure.
gee: Logical, indicating if the data is cross-sectional data or longitudinal/panel data. If TRUE, the generalized estimating equation will be used to correct the standard error estimates. The default is FALSE, indicating that the data are cross-sectional.
id: A vector of character, specifying the cluster index in longitudinal data. It is required when gee is TRUE. The length of the vector should be the same as the number of observations.
corstr: A character string, specifying a possible correlation structure in the error terms when gee is TRUE. The following are allowed: independence, fixed, stat\_M\_dep, non\_stat\_M\_dep, exchangeable, AR-M and unstructured. The default value is exchangeable.
unequal_interval: Logical, indicating if age and period groups are of the same interval width. The default is set as TRUE.
age_range, period_range: Numeric vector indicating the actual age and period range (e.g., 10 to 59 years old from 2000 to 2019).
age_interval, period_interval, age_group, period_group: Numeric values or character vectors indicating how age and period are grouped. age_interval and period_interval are numbers indicating the width of age and period groups respectively. age_group and period_group are character vectors explicitly listing all potential age and period groups. Either age_interval(period_interval) or age_group
(period_group) have to be defined when unequal_interval
is TRUE.
...: Additional arguments to be passed to the function.
A list containing: - model: The fitted generalized linear model.
intercept: The overall intercept.
age_effect: The estimated age main effect.
period_effect: The estimated period main effect.
cohort_average: The estimated inter-cohort average deviations from age and period main effects.
cohort_slope: The estimated intra-cohort life-course linear slopes.
int_matrix: A matrix containing the estimated coefficients for age-by-period interactions.
cohort_index: Indices indicating different cohorts.
data: Data used for fitting APC-I model.
# load package library("APCI") # load data test_data <- APCI::women9017 test_data$acc <- as.factor(test_data$acc) test_data$pcc <- as.factor(test_data$pcc) test_data$educc <- as.factor(test_data$educc) test_data$educr <- as.factor(test_data$educr) # fit APC-I model APC_I <- APCI::apci(outcome = "inlfc", age = "acc", period = "pcc", cohort = "ccc", weight = "wt", data = test_data,dev.test=FALSE, print = TRUE, family = "gaussian") summary(APC_I) # explore the raw data pattern apci.plot.raw(data = test_data, outcome_var = "inlfc",age = "acc", period = "pcc") ## alternatively, apci.plot(data = test_data, outcome_var = "inlfc", age = "acc",model=APC_I, period = "pcc", type = "explore") # visaulze estimated cohort effects with bar plot apci.bar(model = APC_I, age = "acc", period = "pcc", outcome_var = "inlfc") # visaulze estimated cohort effects with heatmap plot apci.plot.heatmap(model = APC_I, age = "acc",period = "pcc") ## alternatively, apci.plot(data = test_data, outcome_var = "inlfc", age = "acc",model=APC_I, period = "pcc")
Useful links