The CPANEL Procedure
PROC CPANEL Compared with Other SAS Procedures
The CPANEL procedure models panel data in a way that is comparable to how the HPPANEL and PANEL procedures in SAS/ETS software do this.
PROC CPANEL Compared with the HPPANEL Procedure
The functionality of the CPANEL procedure closely resembles that of the HPPANEL procedure, which is a high-performance procedure. The CPANEL procedure is the next generation of the HPPANEL procedure, and it was developed specifically for SAS Viya. Both procedures are designed to run on a cluster of machines that distribute the data and the computations. Both procedures perform computations in multiple threads.
The CPANEL provides all the functionality of the HPPANEL procedure, and also provides the following:
instrumental variable (IV) regression estimators
dynamic panel estimators, as described by Arellano and Bond (1991) and Blundell and Bond (1998)
Hausman and Taylor (1981) and Amemiya and MaCurdy (1986) estimators, which are hybrids of fixed-effects and random-effects estimators
estimation of the fixed-effects model by first differencing (FD)
Swamy and Arora (1972) estimates of variance components
tables that compare parameters from multiple models and multiple estimation methods
the option to specify a character variable as the cross section ID variable
the option to not specify a time series ID variable when time is not relevant to the estimation
PROC CPANEL Compared with the PANEL Procedure
The CPANEL procedure provides the following functionality not found in PROC PANEL:
instrumental variables (IV) regression estimators
Swamy and Arora (1972) estimates of variance components
support for estimation when the data contain cross sections that consist of only one observation
the option to specify a character variable as the cross section ID variable
the option to not specify a time-series ID variable when time is not relevant to the estimation
The PANEL procedure parameterizes the fixed-effects model by using the last ordered cross section and the latest time point as reference categories. The CPANEL procedure uses a mean-centered parameterization.
The PANEL procedure supports the Da Silva (1975) method for serial correlation, and the Parks (1967) method for first-order autocorrelation. The PANEL procedure also provides alternative estimators of variance such as heteroscedastic and autocorrelation-consistent (HAC). Such functionality is not provided in this release of the CPANEL procedure.
The PANEL procedure offers a wide variety of postfitting analyses, such as model specification tests and tests for unit root. The PANEL procedure is also able to construct utility data sets that contain lagged variables and data reformatted from wide form to long form. This release of the CPANEL procedure is limited in postfitting and data-generating functionality, because the focus for large data sets is primarily on model fitting and comparison.
The CPANEL procedure is specifically designed to operate in SAS Viya and performs computations in multiple threads. The PANEL procedure executes in a single thread on a single machine.
Modeling Only Time Effects
The PANEL procedure and the HPPANEL procedure provide the following MODEL statement options for estimating models that contain time effects and no cross-sectional effects:
FIXONETIME for one-way fixed time effects
BTWNT for between-groups regression, grouped by time value
You can obtain the same functionality with the CPANEL procedure by setting your time variable as the cross section ID variable, and by leaving the time series ID variable unspecified. The following statements perform estimation equivalent to the FIXONETIME option:
proc cpanel data = mycas.a; id time; model y = x1 x2 x3 / fixone; run;