R help - Jan 2008 - model selection method - step() or bic.glm()

Dear R-helpers,
 
I'm considering two methods of selecting a poisson regression model within
R:
1. Using the step() function (stats package) to find the best model by a
stepwise algorithm and AIC
2. Using the bic.glm() function (BMA package) to find the best model by
Bayesian Model Averaging and BIC
 
Are these both reasonable methods for model selection or is one clearly more
appropriate than the other?
 
I understand this is more of a stats question than an R software question,
but I hope someone can help. Thanks in advance.
 
Best wishes,
Des

	[[alternative HTML version deleted]]