@Ajax Epidemiology is just the term for the study of diseases. It is neither weak nor strong per se.
The weak science is the Observational study where a researcher observes things, often stated in secondary data gathering like public health stats or hospital records rather than personally observing them and then tries to correlate them e.g.These people said they ate these things on average and they got these diseases so I am going to conclude that this diet is more likely to ‘be related to’ this disease. There is lots of statistical shenanigans to attempt to ‘prove’ correlation or even causation or people just latch on to results and assume causation based on their own beliefs without the researcher even implying it - like cholesterol deposits are found in all heart attack victims, so it must be a cause or people who eat meat get cancer so meat must cause cancer (strokes pet sheep) .
The strong science is the clinical trial where the researcher actually controls the inputs and measurement of results directly e.g. we will feed these people X and see if Y happens… and then try to explain it. There is obviously a lot of room for error in this latter form too especially in the design of the study, the initial hypothesis (you can create a study and manipulate the analysis to ‘prove’ pretty much anything) but it is much more likely to be able to show causation that the observational study.
The strongest form of clinical study is the double blind or randomized version where some people get the real stuff (actual drug, specific food etc.) and some don’t but the researchers don’t know which until after the study so they can’t bias the results during the test. The ‘double blind’ is that neither the subjects nor the researchers know who’s getting what. This is how drugs are tested. A good study will have a statistically significant number and spread of subjects e.g. tens or hundreds of subjects depending on what you are testing and women, men, ethnicities etc. so you know how those things impact the results. If it is something uncommon you might only need 10-20 subjects but if general to the overall population you might need hundreds.
Many clinical trials in the past were only carried out on white men so it was not known how a drug or phenomenon might affect women or non-white populations differently e.g. women have heart attacks and symptoms differently to men so the ‘clutching your left arm’ thing is often irrelevant for women while jaw pain might be a better indicator.
They can still bias the analysis etc. so it isn’t protection from that but at least the data should be sound and have a ‘control’ group to compare the results against.
After the study, you get the peer review where (hopefully) eminent scientists in the field review and analyze the study and either declare it flawed or good. As we all know this is often driven by the conventional wisdom in a particular field so often it is about confirming existing bias rather than a totally fair review but it is another protection against totally BS studies making it into the established science. It doesn’t stop snake oil sales-persons running tiny trials with no statistical significance, not getting them peer reviewed but then writing a book to make money off it.