BMisc includes miscellaneous functions useful for applied econometrics, with a focus on panel data and distributional analysis. Utilities cover balancing panels, working with distribution functions, computing weighted statistics, manipulating formulas, and supporting staggered treatment adoption settings (e.g., identifying treatment groups and recovering pre-treatment outcomes).
You can install BMisc from CRAN with:
install.packages("BMisc")Or install the development version from GitHub with:
# install.packages("devtools")
devtools::install_github("bcallaway11/BMisc")make_dist creates a distribution function from a vector
of values and their corresponding CDF values.
library(BMisc)
y <- rnorm(100)
y <- y[order(y)]
u <- runif(100)
u <- u[order(u)]
F <- make_dist(y, u)
class(F)
#> [1] "ecdf" "stepfun" "function"
# plot(F)make_balanced_panel drops observations from a panel
dataset that are not available in all time periods.
id <- rep(seq(1, 100, 1), 2) ## individual ids for setting up a two period panel
t <- rep(seq(1, 2), 100) ## time periods
y <- rnorm(200) ## outcomes
dta <- data.frame(id = id, t = t, y = y) ## make into data frame
dta <- dta[-7, ] ## drop the 7th row from the dataset (which creates an unbalanced panel)
nrow(dta)
#> [1] 199
dta <- make_balanced_panel(dta, idname = "id", tname = "t")
nrow(dta) ## now all the observations with missing data in any period are dropped
#> [1] 198get_group identifies the time period in which each unit
first becomes treated. check_staggered verifies that
treatment is absorbing (no de-treatments).
n <- 100
id <- rep(seq_len(n), each = 4)
t <- rep(1:4, n)
## assign units to groups: 0 = never treated, 2 = treated from period 2, 3 = from period 3
g <- rep(sample(c(0, 2, 3), n, replace = TRUE), each = 4)
treat <- as.integer(t >= g & g > 0)
dta <- data.frame(id = id, t = t, treat = treat)
dta$group <- get_group(dta, idname = "id", tname = "t", treatname = "treat")
head(unique(dta[, c("id", "group")]))
#> id group
#> 1 1 3
#> 5 2 3
#> 9 3 2
#> 13 4 0
#> 17 5 3
#> 21 6 2
check_staggered(dta, idname = "id", treatname = "treat")
#> [1] FALSErhs extracts the right-hand side of a formula.
toformula builds a formula from variable name strings —
useful when constructing formulas programmatically.
ff <- y ~ x1 + x2 + x3
rhs(ff)
#> ~x1 + x2 + x3
#> <environment: 0x5823b0801128>
toformula("y", c("x1", "x2", "x3"))
#> y ~ x1 + x2 + x3
#> <environment: 0x5823b0847790>get_list_element extracts the same element from every
entry in a list — a common pattern when results are stored as a list of
lists.
results <- lapply(1:5, function(i) list(est = i * 1.5, se = i * 0.1))
get_list_element(results, "est")
#> [[1]]
#> [1] 1.5
#>
#> [[2]]
#> [1] 3
#>
#> [[3]]
#> [1] 4.5
#>
#> [[4]]
#> [1] 6
#>
#> [[5]]
#> [1] 7.5
get_list_element(results, "se")
#> [[1]]
#> [1] 0.1
#>
#> [[2]]
#> [1] 0.2
#>
#> [[3]]
#> [1] 0.3
#>
#> [[4]]
#> [1] 0.4
#>
#> [[5]]
#> [1] 0.5