search

annakrystalli / OpenConBerlin_ReproHack

MIT License
5 stars 0 forks source link

A general model for the scaling of offspring size and adult size #3

Open schmelling opened 7 years ago

schmelling commented 7 years ago

Project/paper title:

description:

edit with description

participants

resources

github URL:

if a #ReproHack paper

paper URL: http://doi.org/10.1086/589889

data URL: http://dx.doi.org/10.6084/m9.figshare.1094315

code URL: http://dx.doi.org/10.6084/m9.figshare.1094315

wittenator commented 7 years ago

Comments:

schmelling commented 7 years ago

Could use a second setup tutorial for people who have R installed. Just only more line:

install.packages('smatr')
source("analysis.R")
schmelling commented 7 years ago

Tell us a bit about the paper:

A General Model for the Scaling of Offspring Size and Adult Size Authors: Daniel S. Falster, Angela T. Moles, and Mark Westoby

We didn't know about the paper before and we also had no one in the team that is from a related field.
The paper seems to have a solid model for predicting the offspring size by the adult size, but we actually just read 
the abstract and tried to reproduce the results using the code. The paper is loaded with formulas, which is
important, great, and should be included.

Tell us a bit about your team: Nicolas Schmelling: I'm a first year PhD student in synthetic microbiology at HHU Dusseldorf. I mainly work as a computational biologist that tries to answer biological questions from large datasets. Tim Korjakow: I am a first year CS bachelor student and I am currently starting to specialize in statistics and ML. Besides chemistry, these are also my main fields of interest. What happened during the hack? How easy was it to find/download the code and the data?

Getting the data, as well as the the code, was extremely easy, because the provided links enabled one to download one zip-file from figshare with all the resources. After installing one dependency for R ("smatr") the R program automatically downloads and processes the data to output the graphs as pdfs.

Did you need technical skills to reproduce the figures/stats?

The only needed technical skill you need is the ability to run R code and install an additional module/dependency.

How long did it take (in minutes): Getting the data: 2 min Getting the code: 2 min Having the code running: 5 min (just because downloading a file took a while)

What worked? We could reproduce all of the figures that are data dependent. Figure 1 is an overview of the model.

What didn't?

The formulas in the plots weren't generated.

Any feedback for the author?

The author could include one line of code for installing the required packages. We tested both ways, i.e. running the make command and also just using R Studio preinstalled. In both cases we need to install the packages before running the code. That was not a problem for us, but could be for others. Our idea was to just include one line in the README file.

install.packages('smatr')

before the existing line of code

source('Analysis.r')

The author should add auto-labeling of formulas to the R code which generates the figures.

Maybe but the code on GitHub (get DOI with zenodo to make citable) so we could just added the changes.

What do you think you learned? I think this paper shows create how to build a set of R files to reproduce the result figures from scratch and is a great example for other who don't know how to do it. The figure code was fairly easy to understand. However, it could use some more comments. The equation part on the other side needs comments or self-explanatory variable names otherwise you get lost. This is mainly important when you really want to dig into the math. For basic reproduction purposes the code is fine.

How does the experience make you think about reproducibility? Great and easy when done correctly. I never reproduced a figure from a paper that quick. (Comment from Tim): I second this opinion, because I actually anticipated much more work and problems while trying to reproduce this paper. Probs to the author.

dfalster commented 7 years ago

Thanks so much for trying this out @schmelling and @wittenator. Delighted to hear you could reproduce the figures. I will update github repo to incorporate your suggested changes re package installation. The equations in the figures 4-6 were added outside of R so will not be visible in the generated figures. These days I would try to add them, but back then it was beyond me!