0

I am not referring to the F test, which could be misleading. Ok, so I have one set of Data and two models to describe the data. Problem is, the second model has the same exact parameters of model 1, and the only difference is algebraic, as in we add a 1 to some of the terms in the model. That being said, it makes it hard to compare the two models using an F-test, because there are no extra parameters to fix and compare. What I have been doing so far is to run bootstraps and confidence interval analysis on both my models. I have been thinking of using a t-test on the two residual datasets that I get from bootstrapping. What I mean is that the residual obtained in each bootstrap to be listed as a data set 1 and 2, for both models, and then to run a paired t-test on them. But is that going to tell me anything about the better-ness of one model over the other?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Yasmin
  • 11
  • 1
  • Your [approach to a solution](https://meta.stackexchange.com/questions/66377/what-is-the-xy-problem) is not likely to be suitable. Instead describe your problem in more detail. Can you show what the models are? Can you explain your assumptions? Can you make it clear why you're choosing between these models and how the choice will subsequently be used? – Glen_b Jun 17 '18 at 02:52
  • 2
    Well, I am using a 1D Ising model to describe a repeat protein, basically I am writing a partition function, which consists mostly of exponential terms. In the second model I am adding the 1 to each exponential term to account for something physically. It fits better with the data (which we know from experiment). Now I want to compare these two models. I don’t know if that answers your questions. – Yasmin Jun 17 '18 at 03:24

0 Answers0