The importance of version control

Perhaps the most contentious scientific argument of the early twenty-first century concerns whether “data” is plural or singular. To accidentally say “data is” at a prestigious conference is to lob a metaphorical lettuce into the traditionalist audience. Submit a manuscript with “data are” and expect scathing reviews impugning your pet goat’s morels1 from an avant-garde grammatician. Fortunately, there is a consensus on one topic (no mere 97% here), the chironomid data from Lake Żabińskie are plural.

Version One – January 2015

The published version of the chironomid stratigraphy which has not been archived and is now acknowledged by the corrigendum to be incorrect. None of the archived files match it.

Version Two – January 2015

The version of the data archived at in January differed substantially from the published stratigraphy and contained patterns best described as implausible.

Version Three – June 2015

After enquiring about the discrepancies between the archived data and the published stratigraphy, I was emailed a C2 file with a new version of the data. It has five fewer taxa than V2. These are mostly rare, except for Parochlus which occurred eight times in V2 (four times if the unexpected duplicate assemblages are ignored). Fifty-three of eighty-four assemblages are identical. Some other assemblages have large differences, for example 1931 where most of the Criotopus is replaced by Nanocladius branchio.

Version Four – July 2015

This version was uploaded onto in mid-July. It has the same taxa as V3 but gained samples for 1970, 1969, 1966, 1955 and 1941 and has a sample for 1927 instead of 1925. Only four samples have identical assemblages in both versions: 1986 has complete taxonomic turnover!

Version Five – August 2015

V5 was uploaded onto in August. Seventeen samples have different compositions (more than just rounding errors), for example in the sample from 2002, Paratanytarsus had an relative abundance of 27% in V4 but 0% in V5. It is partially replaced by Tanytarsus sp (14%), while all other taxa also increase in relative abundance. V5 is the first version to show the count sums. As expected, they are generally much lower (median 32.5) than the 50 promised by the paper.

Version Six – September 2015

Another month, another version of the Lake Żabińskie chironomid stratigraphy. This version is mostly the same as V5 rounded to one decimal place, but in 1931 Cricotopus loses 2% and Limnophyes gains 2%.

Version Seven – October 2015

V7 is not rounded to one decimal place as V6 was, nor is it identical to V5. The largest change from V6 is in the sample from 1940, in which 12% Criotopus appears, with all other taxa losing one or two percent.
V7 also includes what are purported to be the raw counts. They have non-integer (or half integer) values which are impossible. I suspect the counts are back-calculated from percent data and then the percentages recalculated.

Version Eight – October 2015

V8 only includes count data which are identical to V7 counts to four decimal places!

Version Nine – November 2015

V9 was uploaded onto NOAA in November. The count data in the excel file are identical to the counts in V7 and V8, those in the text file have been rounded. The percent data are V7 rounded to one decimal place.


Different versions of the chironomid data for 1940

Version ten?

It is not clear that the version of the data described in the corrigendum is the same as V9. The corrigendum details the taxon inclusion rules

Taxa not included in the transfer function or not found in 3 percent in at least 3 samples were removed. In consequence, the number of taxa was reduced from 76 to 50.

Figure 1 caption

Four taxa (Brillia, Eukiefferiella, Krenopelopia and Stictochironomus) present in only 3 samples are not displayed.

However, in V9 (which does have 50 taxa), these four taxa have only one or two occurrences.

Of course, I understand that most people have multiple versions of their data: the initial low resolution counts; taxonomic revisions; corrections of transcribing errors. But this should all be sorted out before publication. Post-publication it should be easy to identify the correct file. To archive an incorrect version once is unfortunate; to archive six different versions (none of which match the published stratigraphy) is beginning to look like carelessness.

Note, all the archived data versions2 have file creation dates that post-date publication.

  1. Admit it, if the morels weren’t fly-blown, your amoral goat would have eaten them (and then would be amorel). 
  2. Data and code are archived on github 

About richard telford

Ecologist with interests in quantitative methods and palaeoenvironments
This entry was posted in Peer reviewed literature and tagged , . Bookmark the permalink.

5 Responses to The importance of version control

  1. chaamjamal says:

    there’s datum and then there are data

  2. Pingback: The missing rare taxa at Żabińskie | Musings on Quantitative Palaeoecology

  3. Pingback: Replicating the Lake Żabińskie reconstruction | Musings on Quantitative Palaeoecology

  4. Pingback: The Humpty Dumpty theory of palaeoecology | Musings on Quantitative Palaeoecology

  5. Pingback: why would anyone not trust the author???? | Musings on Quantitative Palaeoecology

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s