Using the accCost() function in R: questions regarding the transition layer

I am trying to use the accCost() function from the gdistance package in R. I load my cost raster using the raster()function from the raster package.

According to the documentation, I need first to convert my raster object to a transitionLayer object:

tr_cost <- transition(cost,transitionFunction=mean,8)

My results seem to be very sensitive to the way I create tr_cost with this function. According to the parameters I put, I can end up with many cells with a value of 0 where I know it should be something else (my raster is the land of a country, surrounded by NA's). But I must admit I do not quite understand how the two parameters transitionFunction and directionswork.

For transitionFunction, the reference manual does not provide an exhaustive list of the built-in functions and how they work (this page explains a little, but it's still not clear for me). Regarding directions, I found that changing from 16 to 8 increases the number of cells with no data.

I thus have the following questions:

1. What does the transitionFunctioncorrespond to?


2. Is there a transition function that is best suited to deal with accumulated cost?


3. What are the pro's and con's of using 16 directions instead of 8?

Thank you for your help!