p-value for univariate analysis

Question

So I am reading an article in which it has a table showing p-values and hazard ratios for what they say is "Univariate and multivariate analyses of clinical and treatment factors potentially associated with grade 2 toxicity".

However, I might be mistaken, but I thought univariate analysis was on one variable alone, and not in combination with anything else. So can you even give a p-value in a univariate analysis that makes sense in this case ? I mean, if you compare to some kind of mean etc., then maybe, but that doesn't seem to be the case here. Have they maybe mistaken univariate with bivariate, or am I missing something here ?

Answer 1

I agree with you that that is what "univariate" should mean. But it is quite common for people to use "univariate" in the sense of "a single independent variable". Given the rest of the sentence that you quote, that is what was done here.

However, it is possible to have a p-value in a truly univariate case. For instance, a one-way chi-square test is on a single variable and tests a null hypothesis about the distribution of the results. For instance, you could roll a die 1000 times and test the null that the frequencies of each result are equal.

Answer 2

It is possible to talk about p-values in univariate contexts. For example, if you want to compare whether an individual belongs to a population, you may want to measure some variable and contrast with the population expectations, resulting in a p-value to test the null hypothesis "the individual does actually belong to that population"

In this context, there is no big difference between measuring a single variable or many