pggls function

General FGLS Estimators

General FGLS estimators for panel data (balanced or unbalanced)

pggls(
  formula,
  data,
  subset,
  na.action,
  effect = c("individual", "time"),
  model = c("within", "pooling", "fd"),
  index = NULL,
  ...
)

## S3 method for class 'pggls'
summary(object, ...)

## S3 method for class 'summary.pggls'
print(
  x,
  digits = max(3, getOption("digits") - 2),
  width = getOption("width"),
  ...
)

## S3 method for class 'pggls'
residuals(object, ...)

Arguments

• formula: a symbolic description of the model to be estimated,
• data: a data.frame,
• subset: see lm(),
• na.action: see lm(),
• effect: the effects introduced in the model, one of "individual" or "time",
• model: one of "within", "pooling", "fd",
• index: the indexes, see pdata.frame(),
• ...: further arguments.
• object, x: an object of class pggls,
• digits: digits,
• width: the maximum length of the lines in the print output,

Returns

An object of class c("pggls","panelmodel") containing: - coefficients: the vector of coefficients,

• residuals: the vector of residuals,

• fitted.values: the vector of fitted values,

• vcov: the covariance matrix of the coefficients,

• df.residual: degrees of freedom of the residuals,

• model: a data.frame containing the variables used for the estimation,

• call: the call,

• sigma: the estimated intragroup (or cross-sectional, if effect = "time") covariance of errors,

Details

pggls is a function for the estimation of linear panel models by general feasible generalized least squares, either with or without fixed effects. General FGLS is based on a two-step estimation process: first a model is estimated by OLS (model = "pooling"), fixed effects (model = "within") or first differences (model = "fd"), then its residuals are used to estimate an error covariance matrix for use in a feasible-GLS analysis. This framework allows the error covariance structure inside every group (if effect = "individual", else symmetric) of observations to be fully unrestricted and is therefore robust against any type of intragroup heteroskedasticity and serial correlation. Conversely, this structure is assumed identical across groups and thus general FGLS estimation is inefficient under groupwise heteroskedasticity. Note also that this method requires estimation of $T(T+1)/2$

variance parameters, thus efficiency requires N >> T (if effect = "individual", else the opposite).

If model = "within" (the default) then a FEGLS (fixed effects GLS, see Wooldridge, Ch. 10.5) is estimated; if model = "fd" a FDGLS (first-difference GLS). Setting model = "pooling" produces an unrestricted FGLS model (see ibid.) (model = "random" does the same, but using this value is deprecated and included only for retro--compatibility reasons).

Examples

data("Produc", package = "plm")
zz_wi <- pggls(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
               data = Produc, model = "within")
summary(zz_wi)

zz_pool <- pggls(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
                 data = Produc, model = "pooling")
summary(zz_pool)

zz_fd <- pggls(log(gsp) ~ log(pcap) + log(pc) + log(emp) + unemp,
               data = Produc, model = "fd")
summary(zz_fd)

References

\insertRef IM:SEUN:SCHM:WOOL:99plm

\insertRef KIEF:80plm

\insertRef WOOL:02plm

\insertRef WOOL:10plm

Author(s)

Giovanni Millo