I’m quite new to MRS, and carefully read the LCModel manual before initiating my analyses. As far as I understood, the stantard deviations (%SD) obtained per metabolite should be low enough to ensure reliability of the obtained metabolite concentrations, but I wonder if these are relevant as well for metabolite ratios.
Thank you very much for your help!
That is a really excellent question. In theory, yes, the appropriate way would be to propagate the uncertainty of the numerator and the denominator. This isn’t at all trivial for non-linear functions, especially when the variables and their uncertainties are correlated, which is certainly the case in linear-combination modeling. Since the estimated uncertainties for the reference compounds (water or tCr) are usually small, the error incurred from not properly propagating them is probably also small.
NB: For water-scaled concentration estimates, the uncertainties of all contributing variables need to be propagated. In practice, that is also rarely done for quite uncertain parameters like metabolite and water relaxation times, in part because the corrections are non-linear and analytic expressions become unwieldy very quickly. @rinstrella has demonstrated the use of Monte Carlo simulations to circumvent this problem in his ISMRM 2023 abstract (mirasmart.com).
Thank you very much for your response! Therefore, I understand that uncertainties are also relevant when working with ratios, but (in general) following the uncertainty of the numerator would be enough.
Thanks again! Best,