This is debatable as roofs are normally well insulated for residential building. What kind of a structure are you working on ? I have seen steel structures that are open (like in process industry), where temperature gradient is applied based on environmental data. Normally, for temperature effects, with steel you need to be cautious about not to provide more than one braced bays in any direction or else your members between two braced bays will have a restraining force due to temperature. That is the philosophy. What is your structure and why do you want to evaluate your structure for temperature loading ?