| Type: | Package |
| Title: | Inverse Probability Weighting under Non-Monotone Missing |
| Version: | 0.1.0 |
| Date: | 2022-01-01 |
| Maintainer: | Andrew Ying <aying9339@gmail.com> |
| Description: | We fit inverse probability weighting estimator and the augmented inverse probability weighting for non-monotone missing at random data. |
| License: | GPL-2 | GPL-3 [expanded from: GPL (≥ 2)] |
| Depends: | R (≥ 4.0), lava, nloptr, numDeriv |
| RoxygenNote: | 7.1.2 |
| Encoding: | UTF-8 |
| NeedsCompilation: | no |
| Packaged: | 2021-12-19 06:19:54 UTC; andrewying |
| Author: | Andrew Ying [aut, cre], Baoluo Sun [ctb] |
| Repository: | CRAN |
| Date/Publication: | 2021-12-20 15:42:10 UTC |
This function prepares necessary list of information for fitting IPW or AIPW
Description
This function prepares necessary list of information for fitting IPW or AIPW
Usage
PS(data, R, pattern)
Value
returns a list containing the following components:
gamma_para |
the fitted values for the first stage regression |
gamma_IF |
the influence functions of the fitted values for the first stage regression |
sigma_star |
the positivity parameter used in the link function |
prop |
the propensities of fully observed for each R = 1 subject |
This function computes our link function that respects the positivity
Description
This function computes our link function that respects the positivity
Usage
link_func(x, sigma_star)
Value
returns results our link function
This function computes the gradient of our link function that respects the positivity
Description
This function computes the gradient of our link function that respects the positivity
Usage
link_func_grad(x, sigma_star)
Value
returns the gradient of our link function
This function vectorizes link_func_grad
Description
This function vectorizes link_func_grad
Usage
link_func_grad_vec(x, sigma_star)
Value
returns a vector of link_func_grad results
This function vectorizes link_func
Description
This function vectorizes link_func
Usage
link_func_vec(x, sigma_star)
Value
returns a vector of link_func results
Fitting IPW or AIPW Estimators under Nonmonotone Missing at Random Data
Description
nmm_fit is the main function used to fit IPW or AIPW estimators under nonmonotone missing at random data
Usage
nmm_fit(
data,
O,
AIPW = FALSE,
formula = NULL,
func = NULL,
weights = NULL,
...
)
Arguments
data |
a data.frame to fit |
O |
missing indicator |
AIPW |
indicator if fitting augmented IPW |
formula |
optional formula specified to fit |
func |
optional fitting function, currently support 'lm' and 'glm' |
weights |
optional weights used in the estimation |
... |
further arguments passed to func, e.g. family = 'quasibinomial' for glm |
Value
NMMIPW returns an object of class "NMMIPW". An object of class "NMMIPW" is a list containing the following components:
coefficients |
the fitted values, only reported when formula and func are given |
coef_sd |
the standard deviations of coefficients, only reported when formula and func are given |
coef_IF |
the influnece function of coefficients, only reported when formula and func are given |
gamma_para |
the first step fitted valus |
AIPW |
an indicator of whether AIPW is fitted |
second_step |
an indicator of whether the second step is fitted |
second_fit |
if second step fitted, we report the fit object |
by_prod |
a list of by products that might be useful for users, including first step IF, jacobian matrices |
Examples
n = 100
X = rnorm(n, 0, 1)
Y = rnorm(n, 1 * X, 1)
O1 = rbinom(n, 1, 1/(1 + exp(- 1 - 0.5 * X)))
O2 = rbinom(n, 1, 1/(1 + exp(+ 0.5 + 1 * Y)))
O = cbind(O1, O2)
df <- data.frame(Y = Y, X = X)
fit <- nmm_fit(data = df, O = O, formula = Y ~ X, func = lm)
This function prepares the user-provided data into the correct format
Description
This function prepares the user-provided data into the correct format
Usage
nmm_preprocessing(data, O)
Value
returns a list containing the following components:
R |
the indicator of the missing pattern, where R = 1 is reserved for fully observed |
pattern |
a matrix recording the pattern |
This function computes propensities of entries being fully observed
Description
This function computes propensities of entries being fully observed
Usage
prop_compute(para, work_df, R, selection, sigma_star, pattern)
Value
returns propensities of being fully observed
This function returns the residuls after fiting a func regression, where func is the user-specified function
Description
This function returns the residuls after fiting a func regression, where func is the user-specified function
Usage
regress_fit(
coefficients,
regress_func,
formula,
data,
weights,
R,
sigma_star,
pattern,
...
)
Value
returns the residuals of the regression by the user-specified function
Summarizing IPW or AIPW Estimators under Nonmonotone Missing at Random Data
Description
summary method for class "NMMIPW".
Usage
## S3 method for class 'NMMIPW'
summary(object, ...)
## S3 method for class 'summary.NMMIPW'
print(x, ...)
Arguments
object |
an object of class "NMMIPW", usually, a result of a call to NMMIPW. |
... |
further arguments passed to or from other methods. |
x |
an object of class "summary.NMMIPW", usually, a result of a call to summary.NMMIPW. |
Details
print.summary.NMMIPW tries to be smart about formatting coefficients, an estimated variance covariance matrix of the coefficients, Z-values and the corresponding P-values.
Value
The function summary.NMMIPW computes and returns a list of summary statistics of the fitted model given in object.
Examples
n = 100
X = rnorm(n, 0, 1)
Y = rnorm(n, 1 * X, 1)
O1 = rbinom(n, 1, 1/(1 + exp(-1 - 0.5 * X)))
O2 = rbinom(n, 1, 1/(1 + exp(+0.5 + 1 * Y)))
O = cbind(O1, O2)
df <- data.frame(Y = Y, X = X)
fit <- nmm_fit(data = df, O = O, formula = Y ~ X, funct = lm)
summary(fit)