Bayesian Weibull Analysis vs. Weibayes

The Bayesian Weibull analysis capabilities built into BRASS should not be confused the Weibayes method. While both BRASS and Weibayes allow recognize that the past may be an indicator of the future, the Bayesian Weibull analysis features provided by BRASS far exceed the intent and capability of those of Weibayes.

BRASS does not require you to pick a 'best' number for the shape parameter:
The idea behind Weibayes is to incorporate engineering knowledge regarding the aging behavior of your existing products (improving, constant, deteriorating reliability behavior) in the analysis of the product being tested. The engineering knowledge takes the form of a single number, namely an estimated value for the Weibull shape parameter.

The fully Bayesian analysis procedures in BRASS consider the shape parameter to be uncertain, much like the scale parameter. Consequently, if you are not sure about the value of the shape parameter, you can specify the knowledge about the aging behavior in the form of an uncertainty distribution. Doing so prevents you from having to specify numbers that imply too much precision.

The fully Bayesian analysis procedures in BRASS consider the statistical uncertainty about both the shape and the scale parameters, which is shown here in the form of a joint density distribution. BRASS automatically converts these distributions into estimates for the relevant reliability measures as a function of time, plus the corresponding uncertainty bounds.

BRASS does not require you to 'pick' a number for the shape parameter:
Rather than forcing you to specify a number or distribution for the shape parameter, BRASS allows you to directly incorporate the data obtained from your earlier tests into the analysis, and simultaneously analyze the old and the new data, assuming that the shape factor for both data sets are the same, while the scale parameters are independent.

BRASS allows you to use information about the scale parameter as well:
The reliability adjustment model feature in BRASS allows you to not just link the shape parameters of the analyses, but also has the option to propagate information about the scale parameter from one product to another. As an example, you could link the two reliability models based on the assumption that the failure rates are 'approximately the same', which could for instance be translated to a failure rate adjustment by a factor somewhere between 0.8 and 1.3. Similarly, you could model expected increases or reductions in the failure rate, or alternatively perform time-based adjustments.

In summary, since the analysis procedures in BRASS are actually Bayesian, far more powerful reliability models can be constructed, that help you make the most of your reliability data, without having to resort to unrealistically strong assumptions about the parameters in your model.