But I do not endorse about 75% of the “profession” - which doesn’t change my original statement that 833 is a tiny sample for a national poll .and that a “subset” is only a subset when it’s a subset ..and when the subset is the entire set, it’s not a subset anymore.
Two concepts you can’t wrap around .
But data based on a subset has no more statistical validity than the size of the subset. Those larger national polls you are referencing are typically only interviewing Republicans a third of the time. Thus, the Republican-specific results of those polls are only as statistically valid as that smaller subset of Republicans would suggest.
Again, in that context, this is actually a relatively large sample of GOP-only voters.