Repeating Crumley Experimental Data on FigShare!

It took me a while, but I finally got my data from the Repeating Crumley set of experiments up on FigShare. Of course, I didn’t call it Repeating Crumley there since that experiment has no context. You can download the data sets and figures here:

Repeating Crumley FigShare Data and Figures

The interesting thing about this site is that your data remains your data but is openly accesible. Each upload is given a citation. Mine is:

Salvagno, Anthony; Koch, Steven J; Salvagno, Anthony (2012): Repeating Crumley: Tobacco Seed Growth in D2O. figshare.
Retrieved 22:15, Mar 02, 2015 (GMT)

Also there is a little section for social media promotion and it tracks the page analytics for you as well. That is pretty neat!

It did take me a while to get this up on FigShare unfortunately. For a while I couldn’t log in with Twitter, Facebook, or Google. I was asked repeatedly to sign in while trying to upload stuff. Eventually I just gave up and created a new account. Then when I tried to upload my first images, the site was unresponsive. I did spend a considerable amount of time organizing my resources so it would be presentable on FigShare, so that added to the mess a bit. And I also spent some time collecting links from my notebook to incorporate with the data so all experimental information can be collected. I also monified my active experiments page a little so some of the links are easier to navigate.

Despite how time consuming this first run was, it was definitely worth it.






  • Steve Koch
  • Steve Koch

    AWESOME for using FigShare!!!  OK, Now I have more criticism.  Excel is not sufficient for making graphs.  Google sucks even worse for it.  The statistics programming suite, R, is the best choice for us, and you and Alex and all future students need to learn to use it.  You will love it after a few hours of playing around.  It gives you full control of output, though admittedly it’s complicated at first.  But once you get a handle, you will realize how good it is.  And programmatic control of graph output is so so much easier.  Plus, R is a powerful language and will probably lead you to using R for some analysis (instead of LabVIEW or Google spreadsheets).  In order to give you a jump start, I shared a collection with you on Google, with some code for producing the attached figure.  The shared collection link is: 

    I also created a document with some short instructions on how to get started.  It will still require some trial and error, I’m sure, but hopefully you’ll get started a lot quicker than I did.  I’ll paste the instructions below this text too.

    *** Please try this as soon as possible. Once you have the figure recreated, adjust it however you’d like and then please upload the new version to FigShare.  Thanks!

    Instructions pasted from Google Doc: 

    Instructions for recreating this figure in R1.  Download R. Download the latest version (be sure of 32-bit or 64-bit operating system).2.  Download the .r code from Google collection (RC Graph manual.r).  You can view this code with Notepad++. I recommend downloading notepad++, because it will color-code the R code for you.3.  In “R”, change your directory with the Fil->change dir… option.  change directory to the directory where you have the *.r code.4.  Run the script by typing source(“RC Graph manual.r”)  (See the screenshot PNG)That should recreate the figure for you and save the PNG file.  Now you are ready to adjust the figure, which requires learning some R.  Here are good links:* R Tutorial (awesome): (I am linking to a specific page, but the whole book is awesome, see the contents links on the right):* Google’s R style guide: notes for setting up R can be found on this mindmap:

  • Steve Koch

    I made some progress on how to add “error bars” using binomial confidence intervals.  I think we can do it approximately correctly now and when you get back, we can discuss.  To do exactly correctly (what kind of statistics to use) is complicated and I don’t know.  I’ve attached a graph of RC3 DDW compared to DI with 80% confidence intervals using bayesian binomial statistics (which I don’t fully understand).  But I think this kind of graph is what we’ll want, in order to demonstrate significant (or not significant) differences in growth rate / germination rate between DDW and DI.  My notes are scattered, here are some links:

    * Explanation of graph on Google Docs collection:

    * Notes from yesterday on github:

    * Notes from today on github: