Mean raw is simply instances of an event divided by opportunities for the event to occur.  This is the statistic that other tools display.

Mean Bayesian takes into account the distribution of the stat across all players.  For a player you've never seen before, it will show the average value of the opponents in your database.  As you gather more data on a player, Poker Sleuth updates the value using Bayes' Rule.  When you have a lot of data on a player, the raw mean and Bayesian mean are the same (i.e., the player's true value).

The Bayesian credible interval is a range of values that are likely to contain a player's true statistic.  For a player you've never seen before, it will show a range that contains 95% of the opponents in your database.  As you more data on a player, Poker Sleuth updates the ranges using Bayes' Rule.  The range will gradually narrow until both the low and high ends of the range are the same (i.e., the player's true value).

For both of the Bayesian values, Poker Sleuth needs to have had enough time to parse many of your hand histories to build up a profile of your typical opponents.