How use F test to compare models with data points using R

Question

[Bioinformatics1]

I compared experimental data with the nonlinear inhibition kinetic growth models and added rmse in my manuscript.

Now the reviewer asked me F test and R square.

It would be important to show the mean square error of the models used to assess the specific growth rates as a function of substrate concentration to demonstrate how better Aiba et al.'s model fitted the experimental data. Also, the authors should present the F-test and R-sqr of the models

I fit data point to one of models:

mE<-nlsLM(mu~a*(exp(-S/c)-exp(-S/b)),data=df,
        start=list(a=0.07,b=5.74,c=5.53),
        algorithm="port")

P value is larger than 0.05, but the correlation between this model and the experimental data points is more than 90%

How do I explain to the reviewer, I use AIC instead of Rsquare? Does S/he ask me the correct request to measure R?

I use the ANOVA code to compare models. The output is empty for P and F. Why?

Now I have only AIC and rmse but no F value

When I fit the model to data point the p values are not significant. How should I interpret this?

Can you introduce a resource to explain in a simple way?

Thanks

Answer 1

[M.O.L.S]

In R there is the nls function for nonlinear least squares, but I have not seen the nlsLM function. Are you using a special library?

When you create a linear model using lm, the adjusted R squared and Multiple R- squared values are given after using the summary function on the linear model.

You can then use the anova function on the linear model to get the F value. There seems to be an F test used in this anova function.

If you want the p-values from the model you could use broom::tidy() on your model. You can get the AIC by using broom::glance() on your model .

[Bioinformatics1]

Thanks, I found this function on youtube. nlm did not work for my model. I think , it was "minpack.lm" library