nceas / codyn Goto Github PK

Directional change across windows of increasing time lags. Collins et al. 2000

Many related taylor functions.

The file R/codyn-package.R needs to be filled out

The meanrank() function is supposed to take a data frame and the names of the year, species, and abundance variables as input. But if a data frame is passed in that uses non-standard variables, the mean rank shift calculation results in all values as NaN. It appears no attempt is made to handle the non-standard column names. I will commit a test that reproduces this bug, which can be run with the test() command.

Many files have issues with ROxygen. Fix syntax and ensure it all runs with document().

The vignette "codyn: An R package..." contains placeholder text

Working through the rate_change code, need to get it to the state where input is longform and there is the opportunity for replication. Happy to help with that but I'm not totally sure what the test statistics should be -- should I be extracting the coefficient and p value for the lag? a/o R squared? I think we'll probably want the output to be a dataframe with the relevant test statistics paired with the replicate, but I'm not totally sure what they should be, will get unwieldy to have it be the full model summary. Thoughts?

Turnover metrics

• appearance, disappearance, both
• crash metrics
• Beautify / refactor functions
• net change between years

• timevar -> time
• speciesvar -> species
• abdvar -> abundance

Need full tests for variance ratio metrics, including for functions:

• calVR
• calVR2
• getRand
• calNullVR
• nullVRCI
• calVRrealnull
• VR

The current function names are good, but might be improved for clarity and intuitiveness. By checking this before the package is released, we will make it easier for people to understand each function and its purpose. We should also check if these function names are already in use in other CRAN packages. The current function names from the NAMESPACE export file to be checked are:

• VR
• calVR
• calVR2
• calVRrealnull
• calnullVR
• genRand
• meanrank
• nullVRCI
• pairwise_turnover
• timelag
• synchrony
• synch_onerep
• calComDat -> rename to transpose_community()
• calComTS -> deleted function

Check each off once we have agreement to keep the function name.

• timevar -> time
• year -> time
• speciesvar -> species
• abdvar -> abundance
• data1 -> data
• rep -> replicate

The collins08.csv has multiple entries for the species asclvird in all years and replicates, and for the species desmilli in the unburned replicate for years 1994 and 1998. This is causing an indexing error because all our functions require unique abundance values for a species within a year and replicate.

Syd - want to have a look at the collins08.csv and figure out the data cleaning issue? Note that I moved it to the inst/extdata folder

Even when I select years with unique values, this aspect of the meanrank code gives me trouble with collins08.csv but not kn_001d.csv, still trouble shooting:
commonspp <- Map(getintersect, y1, y2, dataname = species)

Need someone to review the current tests for community stability/suggest others.

add a check in the wrapper of each function to make sure that the names passed are the names present

Need tests for synchrony metrics. I added a stub test file, but the meat of the test needs to be completed.

Need tests for turnover metrics.

Need tests for timelag metrics.

• add the equation from equation 1
• reproduce figure 3d (partially)
• add rank clocks

A hybrid between the turnover metric and mean rank shift. A metric that combines appearance of species and changes in ranks. Will require testing as well.

• some metric that assigns ranks for missing species and then calculates

need to shift code from the varianceratio function to the temporal_torus_translation_CI function to make it easy to calculate CI either within or across replicates. If across, need to run each iteration of the null model once, average across replicates, and then repeat.

Current timelag metrics file doesn't allow the package to build because of missing variables and potentially other issues. Refactor to allow the package to build.

Total community stability, mean over sd.

• add rate_change example to temporal diversity vignette
• write codyn overview
• make figures larger
• move rank clocks example to temporal diversity indices
• rename vignettes, delete collins08 once material is moved
• flush out gross discussion
• clean up citations

The current rate_change() returns the slope of the rate change data. Some people have requested the ability to get the lagged data themselves. Implement new functions by refactoring get_slope().

Current namespace is not complete. Make sure it is generated correctly and contains all imports.

Can eliminate by using base::dist rather than vegan.

Lauren H indictaed that we could remove all dependencies on reshape to simplify the package. Matt agreed to do so.

Need full tests for taylor metrics.

Need to complete the list of authors.

Need to complete tests for functions in utilities.R.

Grman et al 2010 use phase scrambling to test sig. of their variance ratio. Although they relied on Matlab, it seems that exists in R as boot::tsboot().

Should we include this?

Need tests for rank shifts.

need to do this for almost all of the functions

cool new metric for synchrony, should include

added a bunch, would be nice for others to have a look

Mean rank shift

• pairwise sets of species between years
• only species present across all years

Function names should be consistent. We chose underscores. Rename:

• meanrankshift -> mean_rank_shift
• varianceratio -> variance_ratio

looks like this file was supposed to be deleted? its contents already seem to be in /R