# Robust regression and different datasets in R

I've been using two different packages in R that work very well, but I wanted to know if there is a way to use them simultaneous.

I have datasets that should behave as non-linear function, dependent on some parameters and initial conditions. Given that both linear and non-linear fitting are very strongly affected by outliers, I used the package "robust" to perform robust regression (I can also linearise my data and fit a robust linear model, but that is another issue altogether).

Now, on another set of data I've been using the package FME to fit datasets where there are several global parameters and some experiment specific parameters. The major advantage here is that I can combine different dataset into one cost function.

So, my question is, is it possible to combine different dataset, in a similar way as in FME, in a robust regression framework?

EDIT: For a more clear question. I a deterministic model that explain the non-linear behaviour of my data. This model has two parameters, $s$ and $f_{m}(0)$, The latest is the frequency of the mutant at the beginning of the experiment. I have $n$ replicate experiments where the same $s$ is expected to exist but where the experiment starts with different $f_{m}(0)$. Given the presence and potential strong influence of outliers, I used robust non-linear regression to fit each experiment individually and then compute the average $s$.

I wonder if the is a better way. Given that I assume $s$ to be constant is all experiments, one could fit the whole dataset with one set of parameters ($s$ and $n$ different $f_{m}(0)$). It is even possible that some replicates are considered complete outliers (their behaviour could be not consistent with one $s$ for all).

• Could you describe what "FME" actually does? – whuber Feb 20 '17 at 17:23
• I added FME as an example. The important feature is the ability to combine data from different experiments/initial conditions. – Diogo Santos Feb 20 '17 at 17:39
• Yes: but since your question focuses on such "combination," don't you think we would need to understand what that entails in order to answer your question? Please tell us more specifically what you are trying to accomplish. – whuber Feb 20 '17 at 17:41