ff_panel_cumsum_grouplast: Cumulative stat value from each group from the latest date of...

In FanWangEcon/REconTools: R Tools for Panel Data and Optimization

Description

svr_id are individuals, observed at multiple points; svr_x is date of observation; svr_y is some value observed at each time point. Imagine these are SAT scores for individuals. we want to calculate the sum/avg etc cumulatively at each date for all individuals, but only using their last score.

Usage

ff_panel_cumsum_grouplast(
  df,
  svr_id,
  svr_x,
  svr_y,
  svr_cumsumtop = "y_movingavg_lastestscores",
  stat = "mean",
  quick = TRUE,
  verbose = FALSE
)

Arguments

df	dataframe
svr_id	string name of individual id variable
svr_x	string name of the ranking variable
svr_y	string name of the value variable
svr_cumsumtop	string variable name that stores the cumulative summed variable
stat	string type of statistics to compute, mean, sum, max, min, median
quick	boolean faster algorithm without repeating calculation, quick should be used, slow was initial bad algorithm, results identical
verbose	boolean when quick is FALSE, could verbose which generates slower results

Value

a dataframe with the cumulative summed/averaged etc column up to each row

Author(s)

Fan Wang, http://fanwangecon.github.io

References

https://fanwangecon.github.io/REconTools/reference/ff_panel_cumsum_grouplast.html https://fanwangecon.github.io/REconTools/articles/fv_panel_cumsum_grouplast.html https://github.com/FanWangEcon/REconTools/blob/master/R/ff_panel_cumsum.R

Examples

library(tidyr)
library(dplyr)
library(tibble)
set.seed(12345)
it_N <- 5
ar_it_M <- sample(1:10, it_N, replace = TRUE)
ar_it_M_ID <- sample(1:it_N, it_N, replace = FALSE)
tb_combine <- as_tibble(cbind(ar_it_M, ar_it_M_ID)) %>% rowid_to_column(var = "id")
tb_long <- tb_combine %>% uncount(ar_it_M)
tb_long <- tb_long %>% add_column(xrand = runif(dim(tb_long)[1])) %>% arrange(xrand) %>% mutate(x = row_number())
tb_long <- tb_long %>% arrange(id, x) %>% group_by(id) %>% mutate(rank_l = row_number())
df <- tb_long %>% select(id, x) %>% add_column(y = runif(dim(tb_long)[1])) %>% arrange(id,x) %>% group_by(id) %>% mutate(y = cumsum(y))
print(df)
quick <- TRUE
ff_panel_cumsum_grouplast(df, svr_id='id', svr_x='x', svr_y='y',
                          svr_cumsumtop = 'y_movingmean_lastestscore', stat='mean', quick=quick)
quick <- FALSE
verbose <- TRUE
ff_panel_cumsum_grouplast(df, svr_id='id', svr_x='x', svr_y='y',
                          svr_cumsumtop = 'y_movingmean_lastestscore', stat='mean', quick=quick, verbose=verbose)
verbose <- FALSE
ff_panel_cumsum_grouplast(df, svr_id='id', svr_x='x', svr_y='y',
                          svr_cumsumtop = 'y_movingmean_lastestscore', stat='mean', quick=quick, verbose=verbose)
ff_panel_cumsum_grouplast(df, svr_id='id', svr_x='x', svr_y='y',
                          svr_cumsumtop = 'y_movingmedian_lastestscore', stat='median')
ff_panel_cumsum_grouplast(df, svr_id='id', svr_x='x', svr_y='y',
                          svr_cumsumtop = 'y_movingsum_lastestscore', stat='sum')
ff_panel_cumsum_grouplast(df, svr_id='id', svr_x='x', svr_y='y',
                          svr_cumsumtop = 'y_movingmin_lastestscore', stat='min')
ff_panel_cumsum_grouplast(df, svr_id='id', svr_x='x', svr_y='y',
                          svr_cumsumtop = 'y_movingmax_lastestscore', stat='max')

FanWangEcon/REconTools documentation built on Jan. 21, 2022, 10:28 p.m.