Description Usage Arguments Value Examples
Run lasso estimation on the prepared matrixes
1 2 | estimate_matches(x_mat, y_mat, w_mat, x1_mat, n_near = 5,
reduced = TRUE, n_folds = 10)
|
x_mat, y_mat, w_mat |
matrixes of independent, dependent variables and weights |
x1_mat |
matrixes of independent variables for the prediction sample |
n_near |
number of the nearest observations to derive a random match. If 'n_near' is greater than 'length(match_vector)', minimum out of two is used to create a sample for selecting a random match value. |
reduced |
if TRUE terurns reduced outpur without specific regression details. |
n_folds |
number of folds for cross-validation |
In both reduced=TRUE
and reduced=FALSE
forms, the function returns
a list with elements. In the form reduced=TRUE
only results of matching
are returned:
y0_hat
is the vecrtor of predicted values based on the result of
the lasso regression with nfolds
cross validation and "mse"
measure for identifying minimizing value of lambda. It uses x_mat
,
y_mat
, and w_mat
for running regression and predictind y0_hat
.
y1_hat
is the vetor of predicted values produced using the
y0_hat
regression results and x1_mat
independent variables matrix.
match
is the dataframe with: columns period_1_id
- index of each
y1_hat
value; column y1_hat
- its' value; column period_0_id
-
position of the nearest match from the y0_hat
vector and
column y0_hat
- value of the nearest match
In the not reduced form the list contains more elements:
items x_mat
, y_mat
, w_mat
and x1_mat
from the inputs to the
functions.
lambda_cv
- result of the cv.glmnet
.
fit
- result of the glmnet
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 | library(dplyr)
library(purrr)
library(glmnet)
library(syntheticpanel)
# Run estimate of a fake data
XX <- as.matrix(mtcars[, !names(mtcars) %in% "hp"])
YY <- as.matrix(mtcars[, "hp"])
WW <- matrix(rep(1, nrow(mtcars)), ncol = 1, nrow = nrow(mtcars))
XX1 <- as.matrix(mtcars[1:10, !names(mtcars) %in% "hp"])
# Running simple estimation and returning
a <- estimate_matches(x_mat = XX, y_mat = YY, w_mat = WW, x1_mat = XX1, reduced = FALSE, n_near = 5)
# Extract regression coefficients
a$fit %>% coef()
str(a, max.level = 1)
# Developing a sample bootsrtap vector
perm_example <- purrr::map(1:5, ~ sample(1:nrow(YY), nrow(YY), TRUE))
# Running estimations on every single bootstrap permutation vector
a_boot <-
perm_example %>%
purrr::map(~ syntheticpanel::estimate_matches(
x_mat = XX[.x, ],
y_mat = YY[.x, ],
w_mat = WW[.x, ],
x1_mat = XX1,
reduced = FALSE,
n_near = 5
))
# Exploring the structure
a_boot %>%
str(max.level = 1)
# Accesssing fit of a specific bootstrap iteration
a_boot[[3]]$fit
# Doing the same with each single bootstrap iteration
a_boot %>%
map("fit")
# Extracting coefficinets from each single bootstrap iteration
a_boot %>%
map("fit") %>%
map(coefficients)
# Combine all coefficients in a table
a_boot %>%
map("fit") %>%
map(broom::tidy) %>%
map(~ select(.x, term, estimate)) %>%
map2(.y = 1:length(.), ~ rename_at(.x, vars(estimate), list(~ paste0(., "_", .y)))) %>%
reduce(full_join)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.