Yes. This question shows how hard it is to do good science. We would need at the start a random sample of projector reliability. Any self selected sample of people introduces an unknown amount of bias. Even if we correct for the different numbers of projectors sold by each brand, which would be easy, we would be left with the impossible task of estimating how biased the group of Acer customers was vs. the Epson owners. Good science begins with randomized samples in your data, and that will not occur with any forum topic questions where people choose whether to participate or not.
For example. It might be that owners of less popular projectors tend to be more enthusiastic about their model and less critical of flaws, perhaps because they chose that one partly because it was not ordinary and mainstream. The most enthusiastic among them would participate more in a forum questionnaire. We could not know how much that affected the results unless we compared them in a randomized sample. We simply cannot make accurate conclusions without objective, randomized data. Sorry to be so strident.
On the other hand, my own non scientific study concludes that all the manufacturers are reliable enough.