Hawaiian hoary bats ('ōpe'ape'a; Lasiurus semotus) were captured and tracked back to roosting locations on Hawaiʻi Island. Roost tree metrics were observed and collected from 2018 to 2021. We observed a total of 56 roost trees used by 46 bats (18 female; 25 male; 3 unknown). We examined roost preferences at the tree-level with discrete choice analysis. Discrete model choice sets were developed based on distinct selection events and served as the observational units at each level, that is roost tree and roost stand. The number of choice sets was determined both by the number of unique roost sites to which a bat was tracked and the duration of the sampling period over which it was confirmed at one or more roosts. A “basic” choice set was comprised of one used site and two random sites for each selection event. For bats observed for a short period (<3 days) at only one roost, we produced a choice set limited to only a single selection event. For bats tracked to only one roost but confirmed at that roost on at least three days, we included an additional independent selection event for that roost. An additional selection event was also assigned to bats that returned to the same roost locations during more than one season (Reproductive season = May to September; non-reproductive season = October to April) and/or year. Bats that used multiple roosts were assigned an equivalent number of selection events, and additional events if confirmed at a particular roost on at least three days. The method estimates the probability of specific habitat attributes being used by comparing selected to available but unselected random sites. We modeled day-roost selection at the tree-level with 91 choice sets for 45 (18F, 24M, 3 unknown) unique ‘ōpe‘ape‘a that included the habitat attributes of 55 unique trees. This data file includes data pertaining to random tree metrics including, height, canopy cover, and habitat classification.