Self experimentation could be an excellent source of data. Quite a few "interesting" things cannot effectively be rigorously tested, either because we have no effective placebo, or because they are inhumane (e.g., the vast majority of so-called "rigorous" diet studies are actually based on self-reporting and willful compliance, because you can't keep people in controlled a lab for more than a few weeks). Also, as Ionnadis pointed out, quite a bit of (supposedly rigourous) published research is simply wrong.

So for a lot of things, we have to do with "amateur hour" experiments, whether conducted in a university lab and published in Nature, or conducted at home and published in a blog.

Self-experimentation can easily disprove a "general truth" but cannot really support one due to lack of controls ( and do recall the scientific method is statement about disproving things, not about proving them) However, if you do stumble on a local version of a "general truth", you can then look for the n->inf version more rigorously.

You can look at e.g. the late Seth Roberts for examples; the ideas are interesting, some of his successes have been replicated by enough people; Among these - correlation between Vitamin D intake time and sleep quality (see gwern for a rigorous test of that experiment, that confirms it for n=|{gwern}|=1, and many other less rigorous; Rigorous correlation among flax seed (supposedly omega 3) intake, butter, bacon and response time for n=|{seth}|=1, which I have not published but have verified rigorously for n=|me|=1 about flax seed. And there are a few others. I think some of his ideas are crazy but almost all are interesting.

(Self-reported) success on his Shangri-La diet is ~80%, which is very high for a diet, but obviously not a general truth since it is not 100%. It worked for me once, but not the 2nd time a couple of years later despite following exactly the same protocol. I'd say self-experimentation is much less rigorous, but way more time-and-captial efficient way of arriving at results.