Is there an R-script or equivalent open source code that automates identification of equations of fit for X vs Y data? I have a software package that identifies and ranks the equation(s) of best fit for a single X v Y data set. The software searches a library of several hundred different equation families (both linear and non-linear)
However, it is unable to batch process. This is problematic as I'm looking to produce models of the relationships between flow (Q) and individual water quality measures for up to 50 different analytes/metrics (e.g. pH, Alkalinity, Ca, SO4, Si, TKN, TN, etc) for up to 90 different surface water sites.
It is apparent from some preliminary analysis that there are linear, exponential, power, logistic and other type of relationships.
Hello Clinton,
I could be wrong but I think it will be hard to find an R script ready to go that will search “several hundred” models. You can probably code something for the major models following these two examples: https://www.zoology.ubc.ca/~schluter/R/fit-model/; https://davetang.org/muse/2013/05/09/on-curve-fitting/.
However, I’d suggest you trying other approaches. For examples, we used R in the past to automate hydrological models runs using a shell command within R to open the model software outside R but the model had autorun (once open), autosave, and auto quit (once the run was finished and saved) functions. You could use R to manipulate the data and save it in the proper format and then call your software but that is may not straightforward. In your case, it could be relatively easier if you use AutoIt (https://www.autoitscript.com/site/autoit/) to automate mouse clicks and keyboard strokes within your software. This way you could use AutoIt to open the files for the different sites, flip through the columns of data representing the analytes/metrics, and run your software for each one. Here are a few resources about AutoIt: https://www.autoitscript.com/forum/topic/105755-autoit-book/; http://onlinelibrary.wiley.com/book/10.1002/9783527801954; https://www.amazon.ca/AutoIT-Scripting-Beginners-Rajan/dp/1514144484.
Hope this helps.
Marcos
Hi Marcos,
Thank you for your helpful message. Makes a lot of sense to automate my exsisting software.
Cheers
Clint
HI Clinton, your question caught my attention. Just for curiosity's sake, can you share the name of the software you're using that does this ranked curve-fitting function? I would be interested in checking it out. Thanks!
Hi Catlin,
It is called Curve Expert:
https://www.curveexpert.net/products/curveexpert-professional/
I find it very useful, and the equations can be customised.
Cheers
Clint
Thank you!
Looks like there is a good Python resource called zunzun that will do what you need:
https://zunzun.com/
https://zunzun.com/temp/static_images/AdditionalInfo.html
https://bitbucket.org/zunzuncode/tkInterFit/
Hope this helps
Caitlin
Hi Catlin, thanks a lot for these links! I'm interested to have a play with some of the data I've been working with.
Cheers
Clint
Hi Clinton,
It's been quite a while since you asked this question, but just caught my attention.
We have published an R package called "netSEM" that does what you asked for.
https://cran.r-project.org/web/packages/netSEM/index.html
It fits linear, quadratic, exponential, logarithmic, even change point (two linear models with different slopes) models and rank-orders all the model fits based on adjusted R^2 values. It finds the best models for all the relationships between the variables given within a dataframe and plots these relationships in a pathway model. It's all automated. You can extract the model equations and statistical metrics for the fits, for instance adjusted R^2 and p-values, easily.
There are other functions in the package as well that you may find useful.
Please check the package if you still in need.
Best,
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts