Error propagation in an equation without an analytical solution

Is it possible to formally propagate uncertainties through an equation which does not have an analytical solution? I am an Earth Scientist working on doing a rigorous determination of the age calculated using (U-Th)/He dating, which uses the equation

He = 8*^{238}U(e^{lambda_{238}t}-1)times7*^{235}U(e^{lambda_{235}t}-1)times6*^{232}Th(e^{lambda_{232}t}-1)times^{147}Sm(e^{lambda_{147}t}-1)

We solve for t numerically because the equation cannot be solved for t explicitly. The other variables (e.g., He, 238U, lambda238) are measured or known, each with their own uncertainty. For the sake of argument, let’s say that I know the uncertainty on each value perfectly and all are gaussian. Is there a way to determine the uncertainty in t using "classical error propagation" (e.g., adding in quadrature) without the ability to solve for it explicitly? I know I could build a monte carlo code to figure it out, but I’m wondering if it’s even possible to determine the uncertainty analytically.

Cross Validated Asked by slekce on November 14, 2021

0 Answers

Add your own answers!

Related Questions

ReLU outperforming Softplus

1  Asked on November 12, 2020 by mike-land


Individual sampling weights and percentages

1  Asked on November 6, 2020 by seth-c


Bayesian Likelihood function range

1  Asked on October 29, 2020 by shamm


Zero inflated continuous outcome variables

0  Asked on October 26, 2020 by michaelkyei


How to test paired observations

1  Asked on October 23, 2020 by doug-fir


PCA loadings of correlation matrix

0  Asked on October 20, 2020 by mri


How to model count data with decay

0  Asked on October 15, 2020 by learning-stats-by-example


Ask a Question

Get help from others!

© 2021 All rights reserved.