Statistics is an integral part of hypothesis testing. You can’t do science without statistics. For example, statistics is the basis for the scientific conclusion that smoking increases the risk of lung cancer. When I do a large qPCR expression array, I use statistics as a first pass to see which changes in gene expression are significant. In mouse models, how do you tell if a drug is effective? Statistics. How many mice do we need in each group before we can say that we will see an effect, if it is there? Statistics.
Can statistics give us a false positive or false negative? YES!!! That is why repeatability and consilience is so important in science. If you test a hypothesis using multiple independent tests and methods and you see a statistically significant signal in all of those independent tests then that gives you much more confidence that your hypothesis is correct. Science is always tentative, but that doesn’t mean we hold off on making conclusions until we have 100% knowledge of everything in the universe.