This page has been archived and will not be updated. This is
because it was submitted as part of the publication of
respR in Methods in Ecology and Evolution, and has been
retained unchanged for reference. Any results and code outputs shown are
from respR v1.1 code. Subsequent updates to
respR should produce the same or very similar results.
We have designed respR to be able to explore, process
and analyse any and all aquatic respirometry data, independent
of the system used to collect it or type of experiment. Because of the
unitless nature of the majority of functions in the package, other
respirometry data (e.g. aerobic), or other time series data examining
other variables can be explored and analysed in respR.
The use of simple data structures (numeric vector and data frames)
means there should be a low barrier to entry for anyone not completely
new to R. The unitless nature of the data (requiring only paired values
of numeric time-elapsed, and an oxygen amount in any unit) greatly
reduces the inputs each function requires, and simplifies any analysis
using respR in comparison to other packages (see A
comparison of respR with other R packages). Once the user has
imported and prepared their data to this form (see Importing
your data), data analysis using respR is simple and intuitive.
Both rMR and FishResp centre around processing intermittent-flow, swim tunnel respirometry data with multiple, regularly-spaced replicates, particularly from Loligo Systems equipment. If this does not describe your experiment there is little point in considering these packages, as analysis of other respirometry experiments in them is challenging, if not impossible.
respR is more than capable of processing these
intermittent flow experiments with a little forethought and data
organisation (see
here), and support for this will get better in the coming months.
However, we would encourage users to explore these other packages and
how they work; they are useful options and may contain functionality
that suits your particular analyses or workflows better than
respR.
Importantly, the packages mentioned above only allow manual selection
of data regions over which to determine rates, such as over a specified
time period. This is perfectly acceptable for many analyses, and
respR has this functionality (although in a much more
flexible implementation), but respR also has the
auto_rate() function which identifies linear regions of
respirometry data.
This powerful function allows identification of most linear, minimum and maximum rates in an independent, objective, and statistically robust manner. We would encourage the respirometry community to explore this objective method rather than rely on manual selection, which can leave investigators open to accusations of cherry picking and bias when reporting metabolic rates.
Another R package, LoLinR, can identify linear sections
of non-linear data, athough using a fundamentally different method than
auto_rate(). However, LoLinR is extremely
computationally inefficient, taking literally hours to days to process
typical respirometry datasets. It does appears to perform well, even
after subsampling longer data to the shorter lengths the function can
handle, but auto_rate appears to perform equally well and
can process these data in seconds without modification (see here).
Our other main objective with respR was to provide a
solution for reporting analyses of respirometry data in an easily
reproducible form. See Open
science and reproducibility using respR for an example, but in
summary, an entire respR analysis can be reported in only a
few lines of code. Our careful selection of descriptive function names
and input operators allow - we hope - this code to be readable and
convey easy understanding about what is being done in the analysis, even
without additional comments. Inclusion of a raw data file with this code
would allow anyone to reproduce it easily, and scrutinise each stage. We
hope this makes the job of investigators, reviewers and editors
easier.
We see only a few scenarios in which other R packages might be considered for analysing respirometry data:
LoLinR - If your data is shorter than around 500
datapoints, and you would like to use a different linear detection
method other than auto_rate in respR,
LoLinR can achieve this. Note however, this package is not
respirometry focussed. You would still need to utilise
respR or other solutions to format and import respirometry
data, apply background corrections, and convert the resulting slopes to
particular units. If your data are longer than around 400-500 in length,
LoLinR is not a practical option because of the time needed
to process the data.
rMR or FishResp - If you are doing
intermittent-flow, swim tunnel experiments with regularly spaced
replicates (particularly using Loligo Systems equipment) these are good
options for analysing your data. These data are also able to be analysed
in respR however: see link.
Gas/Air respirometry - We don’t know of any options in R to
analyse air respirometry experiments, and at present respR
is chiefly designed to analyse aquatic respirometry data. However, the
great majority of the exploratory and analytical functions are unit
agnostic. The only step requiring units is conversion of slopes to rates
in the function convert_rate(). While we have not done
this, we imagine air respirometry data could easily be explored and
rates determined, and most experienced investigators in air respirometry
would be able to convert these to units themselves. We plan to add this
support, but as aquatic biologists we have no experience with these
data. Please get in touch
if you can help with this.
respR for your analyses?
In our opinion, you should use respR if you have
any aquatic respirometry data.
We have built respR to accept any and all types of
experiment or data source easily. We have not processed every data
format or variation of experiment, and there may be circumstances we
have not anticipated. So, if for whatever reason some part of
respR does not work for you please do let us know, and we will work
to accommodate it.