** by Gösta
Hägglund**

** Research
Report 1998:3**

Department of Statistics, Stockholm University, S-106 91 Stockholm, Sweden

**Abstract**

A strategy is proposed for correcting inappropriate reference variables when using the factor analysis estimation methods FABIN based on instrumental variables due to Hägglund (1982) and (1998). This is in line with Hägglund (1993), (1994) and (1995). The strategy is intended to be used in situations where one has a certain à priori-conception of the factor pattern to be estimated. At most three non-appropriate reference variables may be involved for cases with an arbitrary number of factors. The strategy is formally based on the concept FABIN-dominant diagonal in Hägglund (1995). Moreover, the strategy is evaluated by means of real data. Here the concept "a most appropriate reference variable" is introduced and defined. For the types of situations and data considered in this report it is concluded that the strategy seems to work well.

**Key words:** Cointegration,
Saddlepoint approximation.

*Last update: 1998-05-14 / KH*