Testing for potential survey bias: the effect of roads, urban areas and nature reserves on a southern African mammal data set