What is my null hypothesis?
It would occur to me, at least if I were the one doing this, and I’m not, that I would want to do multivariate regression calculations with the data to determine a possible relationship first. You did say you’re trying to find a relationship, not just test whether you think the mean amount of water lost equals the mean salinity level. After you come up with a regression equation, whether it be linear or nonlinear, you may then test how accurate you believe your regression equation is, and methods exist for doing so. If you truly are simplifying it to the point of testing if the means are equal, then as stated by another contributor, yes, your null hypothesis (which always contains the condition of equality) would be that they the means (if that’s what you are testing) are equal. You might also test if the mean level of salinity is greater than or equal to the mean amount of water lost (u1>=u2) which would again be your null hypothesis. If you want to test that they are not equal, that
