Hello World!

Welcome to Rstats4ag.org, Statistical Analysis of Agricultural Experiments with R. The current version of this site should still be considered beta quality; there are undoubtedly quite a few grammatical and typographical errors. The explanations are not as clear as they could be, and certainly not as clear as we hope they eventually will be. But rather than wait until everything is "perfect", we are releasing the site into the wild now in hopes that it is useful, and also so that we can get feedback on errors or ways to make it better. The hope is that this site will continually improve to the point that it becomes a valuable resource for graduate students and scientists in the agricultural sciences. Please read the introduction for more information. Over time, we want to add additional contributors and reviewers to the site so that we can increase the scope and utility of the examples provided. If you are interested in contributing in any way, please let us know!

Initial development of this website began just over 1 year ago, but it has really been about 10 years in the making. I thought telling the story of why this site exists would be appropriate for the first blog post here.

In 2003, as I was finishing up my Masters degree and developing a proposal for my PhD, I was trying to analyze some herbicide toxicity data using nonlinear regression. The standard software for such things at the time was SAS PROC NLIN. At the time, there were a few helpful journal articles and other resources available, and after a few days of teeth-gnashing and initial parameter estimating, I had a model functioning and my data analyzed. The next step was creating a figure so that I could present that data during my proposal seminar. I had no idea how to do that; converting results of the analysis into a 'pretty' figure was not included in the resources that helped me run the analysis. I asked a few other graduate students and faculty members, and all of them mentioned other programs like SigmaPlot, Harvard Graphics, etc. But all of those products cost money, far more money than I had access to as a graduate student. I spent a few more days in the computer lab with SAS, and eventually kludged together the glorious figure you see below. Unfortunately, the only copy of that figure I can find came from a corrupted backup disk, so there are some glitches that were not there in the original figure. But I think you can still see the pretty remarkable quality I was able to come up with.

A really poor quality figure

A slightly corrupted version of the best figure I could make given the combination of my skills and software available to me in 2003.

I want to make it clear that this figure isn't necessarily indicative of the quality of SAS graphics at the time; there were probably much better ways to make a pretty figure. But I couldn't find any better instructions online or in print. The very talented statisticians and fellow graduate students I asked for help didn't have any better ideas. The figure above is a product of my relatively novice programming skills and a lack of readily available resources to make it better.

It was pretty obvious to me that the figure above was not going to cut it when it came time to publish my dissertation. So not long after I made the figure above, I started searching for software tools (within my price range) that would allow me to create a publication quality figure. One piece of software kept surfacing during my search: R. There were some major advantages of using R. Since it is a statistical programming language, I could run the analysis and create the graphics in a single program; no switching back & forth as the data/analysis change. And the price was right: FREE! So I installed R, and a few times per month for about a year I would start it up, get completely frustrated by the command line interface and dearth of information on how to use it, and go back searching for other alternatives. It wasn't until about a year of doing this that things began to 'click' and I was able to start doing nonlinear regression in R. And the coolest part was that after I successfully fitted the first nonlinear model, all it took was one more line of code to plot it! A couple more lines of code and I had a figure that I could put into my dissertation. It was a very slow start, but very worth the effort in the long run.

So why this website? Simply, I hope to shorten the learning curve for people who are now where I was 10 years ago. There are certainly far more resources available to students now than there were then. And my colleagues who currently use SAS tell me getting good graphics is much easier than a decade ago. But for those who wish to use the freely available R (particularly students), I wanted to build a freely available resource with examples from my own discipline. I discussed this possibility with Dr. Jens Streibig a year ago, and he was excited to contribute. The two of us have spent the last 2 months assembling data sets, explaining analyses, and building the site you now see. We know there is still some work to be done here, but we decided not to let the perfect become the enemy of the okay. This site is certainly not perfect (and never will be), but we hope the examples here are more relevant to agricultural experiments than many of the examples currently on the web and more accessible than those in textbooks. We hope it will be useful to at least a few people. So please check out the links in the left menu bar, and let us know if you have any input whatsoever; criticism, suggestions, contributions, etc. We hope this is only the beginning!

 

2 Comments

Leave a Reply

Your email address will not be published. Required fields are marked *