Saving and Retrieving Estimates for Start Values

<- Previous: Start Values
Next ->: Retrieval of Start Values

Good start values are essential in reducing total time to convergence in any kind of modelling. For this reason you will want to be able to save estimates and re-use them as start values. For example, suppose you are testing a Normal model against one using a t-distribution. Computer time could be saved by using the estimates from the one model in estimating the second model.

Estimates are saved in the FANPAC global _fan_Estimates after each estimation.   This global can be accessed after the first run and used for a starting point in a second run.

When there are multiple runs in a session, it is important to note that the estimates from each run are stored in columns of _fan_Estimates in the order of the runs. Different runs are likely to contain different numbers of parameters, and their lengths are padded with missing values when appended to _fan_Estimates.

    library fanpac;
    session amzn 'daily Amazon.com';
    setDataSet stocks;
    setSeries AMZN;
    computeLogReturns 251;

    estimate run1 garch(2,1);

    _fan_Start = _fan_Estimates[.,1] | 5;

    estimate run2 tgarch(2,1);

    _fan_Start = _fan_Estimates[1:4,2] | 0 | _fan_Estimates[5:6,2];

    estimate run3 tgarch(2,2);