reg: Unified regression analysis
In KatoPachi/Rkato: My R functions

Description Usage Arguments Details

View source: R/reg.r

This function implements various regression analysis, especially econometrics. This function has some advantages. (1) Unifying augments. By just changing method augment, you can use other regression analyses. (2) Specifying each components of formula. You can easily incorporate this function in a loop system.

1	reg(method, ...)

`method`	a srting of method of the generic function `reg`
`...`	augments which pass on each method.

There are four methods.

`fe` method

Using the package lfe or fixest, you implement the fixed effect, and the panel IV model. In addition to original class, a returned object also has a class fe_felm (lfe pkg) or fe_fixest (fixest pkg). Thus, you can use summary(). You can pass following augments on this method:

y: a formula whose rhs is outcome (y ~ .)
x: a formula whose lhs is covariates (~ x1 + x2)
z: a formula whose rhs is endogenous variable and lhs is exogenous variable (x3 ~ z)
fixef: a formula whose lhs is fixed effects (~ fix1 + fix2)
cluster: a formula whose lhs is cluster variable (~ clust1)
data: a object whose class is data.frame.
pkg: a string of package. Default is "fixest"
se: method of caclulation of se (only fixest pkg). the augment se passes a character scalar: "standard", "hetero", "cluster", "twoway", "threeway" or "fourway." By default if there are clusters in the estimation: se = "cluster"; otherwise se = "standard".

`lpm` method

Using the lm() in the stats package, you implement the linear proability model. To claculate robust standard errors, this method uses coeftest in the lmtest pkg, and vcovHC in the sandwich pkg. If treatment variable is specified, you can estimate this model without covariates and with covariates. A returned object is a list which includes lm result called fit, robust se test called test, and treatment variable formula called treat (if specified). The returned object has lpm class in addition to original classes. If treatment variable is specified, we add RCT class. You can pass following augments on this method.

y: a formula whose rhs is outcome (y ~ .)
x: a formula whose lhs is covariates (~ x1 + x2)
d: a formula whose lhs is treatment variable (~ d).
data: an object whose class is data.frame

`ols` method

Using the lm in the stats package, you implement the linear regression model. If treatment variable is specified, you can estimate this model without covariates and with covariates. A returned object is a list which includes lm result called fit, and treatment variable formula called treat (if specified). The returned object has ols class in addition to original classes. If treatment variable is specified, we add RCT class. You can pass following augments on this method.

y: a formula whose rhs is outcome (y ~ .)
x: a formula whose lhs is covariates (~ x1 + x2)
d: a formula whose lhs is treatment variable (~ d).
data: an object whose class is data.frame

`binomial` method

Using the glm in the stats package, you implement the logit or probit model. A returned object has binomial class in addition to original classes. You can pass following augments on this method.

y: a formula whose rhs is outcome (y ~ .)
x: a formula whose lhs is covariates (~ x1 + x2)
link: character string: "logit" or "probit".
data: an object whose class is data.frame