MULCOM

Authors: Peter Reinhard Hansen, Asger Lunde and Sébastien Laurent

MulCom is a package that provides tools for simultaneous comparison of any number of models. MulCom is designed for forecasting evaluation to analyze whether any among a set of competing models are significantly better than the others, in terms of predictive accuracy. It is however, applicable to a much wider range of problems. It can be applied in any setting where one what to compare the means of two or more populations.

The current implementation of MulCom can do two things. Test for Superior Predictive Ability (SPA) and estimate a Model Confidence Set (MCS). The test for SPA is that of Hansen (2005) which is similar to the reality check by White (2000). The key difference is that the SPA-test is more powerful in many situations. Both the test for SPA and the reality check are implemented in MulCom. The MCS methodology is due to Hansen et al. (2011).

MulCom is a class written in Ox (see Doornik, 2013b), and it is used by writing small Ox programs which create and use an object of the MulCom class. Some knowledge of Ox will be required when using MulCom in this way.

MulCom can also be used interactively in conjunction with OxPack for OxMetrics (see Doornik, 2013a). This is probably the easiest way to use the program, but it requires access to this software. Below we will give examples of applications, both using the console and the interactive version.

MulCom is now integrated into the professional version OxMetrics since version 9.

In case you use the free version of Ox, download MulCom 3.0 here: mulcom.zip

Download the documention here.

Installation:

1. Make sure you have properly installed Ox version 9.00 or later. The MulCom package does not work fully with earlier versions of Ox. Type oxl at the command prompt to check.

2. Unzip mulcom.zip to the ox\packages folder.

3. Read the read.me file for info on the last updates.

4. If Ox has been installed properly, this will allow using the MulCom package from any directory. To use the package in your code, add the command

#import <packages/MulCom/MulCom>

at the top of all files which require it.