Original question: "Majority" is to "plurality" as "minority" is to what?"
I had the exact same question, and studying the answers by others, I came to the insight that the correct answer is actually "smallest subset".
First, I think that a better definition of "plurality" is: the largest subset of a kind. In the example of the OP, 3 weight categories are considered, and each category corresponds to a subset of all people. The plurality is the largest of these, hence the obese.
Now, in most definitions I have seen, the definition of "plurality" is defined as the largest minority, where minority is defined as a subset that's less than 1/2 of all elements considered. However, if we just look at all subsets, and one of them turns out to be a majority (hence more than 1/2 of all considered elements), then it is still a plurality with my preferred definition (there isn't an exception to consider). On the other hand, if more than 2 subsets are considered, and we'd ignore the majority, is it useful to talk about a plurality as the largest of the remaining minorities? In that case, if there is a majority, the largest minority can not be more than 1/4 of all the elements considered (please, check). But how is that relevant? This seems futile and rather confusing.
So, I think that the better definition of plurality is simply the largest subset of a kind, as proposed.
Now, let's look again at the analogy
majority : minority = plurality : unknown.
Consider first that if there are only 2 subsets, they're either both equal in size, and there is no majority, nor a minority (since the definition of majority is that it ought to be more than 1/2, and a minority ought to be less than 1/2). But if there is a majority, there is certainly a minority (the rest), and the plurality equals the majority.
In fact, we saw earlier that the majority, if it exists, always equals the plurality, not just in the case of 2 subsets.
Consider next that if there are more than 2 subsets, then considering majority vs. minority, we're considering at most 1 subset vs. at least 2 subsets. This is an asymmetrical situation. In particular, there can be only at most 1 majority, but there will be more than 1 minority. However, since there are more than 1 minority (and if there is no majority), then looking at the largest minority makes sense and that is the most common situation.
In this case, we can still talk about the smallest minority, which is simply the smallest subset; but it is nonsensical to talk about the smallest majority (you don't need a second term to describe the same unique thing, majority).
Let's translate the analogy to
"subset > 1/2 of elements" : "largest subset" = "subset < 1/2 of elements" : unknown.
We want to solve for unknown. My proposed candidate is "smallest subset". Let's see how this fits.
First, in case of 2 subsets, we see that the relationship on the left expresses the identity, if there is a majority. But if there isn't a majority (both subsets are equal), then we also know that the largest subset equals the smallest subset. In both cases, we can thus write the valid analogy:
"subset > 1/2 of all elements" : "largest subset" = "subset < 1/2 of all elements" : "smallest subset".
Second, in case that there are more than 2 subsets, we need to reconsider the meaning of the analogy expressed.
a.) If there is a majority (unique), then the largest subset is equal to the majority. b.) If there is no majority, then all subsets are minorities, and the largest subset equals the largest minority.
Let's see how we can translate this into the right side terms, mutatis mutandis:
a.) If there is a minority (not unique), and then the smallest subset is equal to a minority. b.) If there is no minority, then TRUE, and whatever.
Note. The b.) leg is seemingly collapsing here. However, if there are more than 2 subsets, it is impossible for there not to be a minority. Namely, in this case, all subsets should be at least 1/2 of all elements (no minority), but with at least 3 sets, this can't be true (unless the subsets are overlapping, but we are implicitly assuming they don't). This means that the b.) leg is TRUE by default (and I just asserted that with "TRUE", but copied the format of the earlier b.) leg for the case of 2 subsets. Consider the definition of an if-then assertion! In particular, the statement "if A then B" is true if A is not true, such as here.
In particular, in the analogy we can use for whatever = "the smallest subset equals the EMPTY set" (this is untrue, but that doesn't matter; it fits the analogy!)
So, surprisingly, for more than 2 subsets, the analogy can include the b.) legs as well! The a.) legs would work with = "smallest subset", if we ignore the slight asymmetry between majority and minority.
Hence, the best candidate for unknown equals "smallest subset", yielding the analogy
majority : minority = plurality : smallest subset.