On the contrary - you come up with a hypothesis and then you test to see if that hypothesis is true.
You make guesses from observations, such as "God strikes down the unworthy" and then you attempt to find worthy people and unworthy people and follow them to see whether the unworthy are striken down in supernatural events at a statistically greater rate than those who are worthy. By using a second set of scientists or clergy who are unfamiliar with your research, you can sort into various forms of unworthiness to see if there is a type bias - sexually deviant, unfaithful, unrepentant, vanity, boastfulness, and others. Your belief that certain unworthiness will result in smiting by a deity is then tested and you review your data.
You may find that God's wrath is not statistically biased towards the unrepentant sinner. Being wrong isn't a problem in science - it's just a path to being right. So, for instance, if you find that your original hypothesis that God strikes down the unworthy is not just incorrect, but backwards. If it seems the virtuous are more likely to get stricken down, and that those of greatest natural virtue are our youth, you can then present this. It may, in fact, then be used to change behavioral patterns and encourage participation in activities. The great researcher into this particular effect, Billy Joel, was instrumental in bringing this research to light, indicating in one of his more widely distributed papers "only the good die young."