Home  /  Resources & support  /  FAQs  /  Specifying the time variable in xtgls

What role does the time variable play in xtgls?

Title   Specifying the time variable in xtgls
Author James Hardin, StataCorp

This depends on the model that you are estimating. Recall that there are three choices for the panel variance structure and three choices for the within-panel structure (see xtgls). Below we show how the values of the time variable are used in the estimation.

panel(iid) corr(indep)

The time variable is ignored. Panels may be unbalanced.

panel(iid) corr(ar1)

The time variable must have values that are a constant difference apart. The time values within a panel must be unique. Panels may be unbalanced.

panel(iid) corr(psar1)

The time variable must have values that are a constant difference apart. The time values within a panel must be unique. Panels may be unbalanced.

panel(het) corr(indep)

The time variable is ignored. Panels may be unbalanced.

panel(het) corr(ar1)

The time variable must have values that are a constant difference apart. The time values within a panel must be unique. Panels may be unbalanced.

panel(het) corr(psar1)

The time variable must have values that are a constant difference apart. The time values within a panel must be unique. Panels may be unbalanced.

panel(corr) corr(indep)

The time values do not have to have a constant difference. The time values within a panel must be unique. The collection of time values must be the same for all panels. The panels must be balanced.

panel(corr) corr(ar1)

The time variable must have values that are a constant difference apart. The time values within a panel must be unique. The collection of time values must be the same for all panels. The panels must be balanced.

panel(corr) corr(psar1)

The time variable must have values that are a constant difference apart. The time values within a panel must be unique. The collection of time values must be the same for all panels. The panels must be balanced.