Results presentation: Outlier probabilities

This page allows you to view the prior and posterior probabilities that particular determinations are outliers.

Accessing this page

This page is accessed from the results selection page by selecting "outlier probabilities".  The option is only available when determinations have been defined with non-zero prior outlier probabilities within the project definition.

Description

A table of all the determinations with non-zero prior outlier probabilities is shown.  For each determination, the lab code is
shown, along with the prior probability of the determination being an outlier (defined by you in the project definition) and
the posterior probability of the determination being an outlier (calculated by BCal during the calibration).

From this information you can assess whether individual determinations in your project are likely to be outliers given all of
the information supplied (in the model, prior and data).  In general, if the prior probability of being an outlier is higher than
the posterior probability, that particular determination is less likely to be an outlier then you originally thought. On the
other hand, if the posterior probability of being an outlier is higher than the prior probability, the determination is more
likely to be an outlier then you originally thought.

Being able to compare the prior and posterior probabilities takes some practice, but it is vital that you resist the temptation
to experiment with a range of prior probabilities.  Prior means just what it suggests.  You must assess your prior
probabilities that samples are outliers on the basis of information available to you BEFORE you obtained the
determinations from the laboratory (factors like `old wood', possible contamination, etc.).  You should never use this option
simply on the basis that you don't like the look of some of the determinations you obtain!
 

Move back to previous pageMove up a level