R/data_filters.R

In GregoryBrownson/EAPR: Empirical Asset Pricing Research Platform

# Implementation of some generic filters used for asset pricing in the past.
#' Filtering data
#' 
#' @importFrom data.table ':=' as.data.table
#' @importFrom lubridate year
#'
#' @export applyFilter
#' @rdname applyFilter
#'
#' @param x An eapr object
#' @param type Filter or list of filters to apply to dataset. Currently, only the filter used in Fama French 1992 is implemented.
#'
#' @return A filtered EAPR object
#' 

applyFilter <- function(x, type) {
    
    valid.filters <- c("ff92")
    if (!all(type %in% valid.filters)) {
        invalid.filters <- type[!(type %in% valid.filters)]
        if (length(invalid.filters) == 1) {
            stop(paste0(invalid.filters, " is not a valid filter!"))
        } else {
            stop(paste0(invalid.filters[1:(length(invalid.filters) - 1)], " and ", invalid.filters[length(invalid.filters)], " are not valid filters!", collapse = ", "))
        }
    }
    
    for (i in 1:length(type)) {
        x <- do.call(paste("filter", type[i], sep = "."), list(x))
    }
    
    return(x)
}

#' Filtering data using methods detailed in Fama French (1992)
#' 
#' @importFrom data.table ':=' as.data.table
#' @importFrom lubridate year
#'
#' @export ff92Filter
#' @rdname ff92Filter
#'
#' @param x An eapr object
#'
#' @return A filtered EAPR object using the same filtering methods used in Fama French (1992)
#' 

ff92Filter <- function(x) {
    # valid.subset <- x$ccm %>% group_by(permno, rebalance_date) %>% mutate(month = month(date), has_Jun = any(month == 6), has_Dec = any(month == 12)) %>% dplyr::filter(has_Jun ==
    # TRUE && has_Dec == TRUE && is_financial == 0) %>% ungroup %>% as.data.table
    
    dt <- x$ccm[, .(date, rebalance_date, permno, log_size, is_financial)]
    dt[, `:=`(has_Jun, !is.null(log_size)), by = list(rebalance_date, permno)]
    dt <- dt[has_Jun == TRUE & is_financial == 0]
    dt[, `:=`(rebalance_date, rebalance_date)]
    
    dt <- unique(dt[, .(permno, rebalance_date)])
    
    x$ccm <- merge(x$ccm, dt[, .(permno, rebalance_date)], by = c("permno", "rebalance_date"))
    
    x$ccm[, `:=`(c("is_financial"), NULL)]
    
    # valid.subset <- x$market.dt %>% group_by(permno) %>% arrange(date) %>% mutate(count = seq(n())) %>% ungroup %>% group_by(permno, rebalance_date) %>% mutate(valid_obs =
    # any(count >= 24)) %>% dplyr::filter(valid_obs == TRUE) %>% ungroup %>% as.data.table
    
    dt <- x$market.dt[, .(date, rebalance_date, permno)]
    dt[, `:=`(rebalance_date, rebalance_date + years(1))]
    dt[order(date), `:=`(count, seq(from = 1, to = .N)), by = list(permno)]
    dt[, `:=`(valid_obs, any(count >= 24)), by = list(permno, rebalance_date)]
    dt <- dt[valid_obs == TRUE]
    
    x$ccm <- merge(x$ccm, unique(dt[, .(permno, rebalance_date)]), by = c("permno", "rebalance_date"))
    
    x$market.dt <- x$market.dt[permno %in% unique(x$ccm$permno)]
    
    cols <- intersect(c("book_market", "asset_market", "earnings_price"), colnames(x$ccm))
    
    return(na.omit(x, cols = cols))
}