eloch216 / photogea Goto Github PK

Will be titled "Working With Extended Data Frames"

There are several places in the vignettes where there could be direct hyperlinks to other vignettes, but instead, just the name of another vignette appears. These instances could be converted to hyperlinks, as explained here: https://stackoverflow.com/questions/34946219/linking-r-package-vignettes

The calculations that determine gm are not documented very well

If an A-Ci curve with one or more Ci < 0 is passed to a fitting function, it will not be possible to find a fit. Make sure this doesn't cause a fatal error. Instead, we can just return NAs for parameters and fits.

There is a lot of duplicated code between fit_c3_aci and fit_c4_aci. It would be nice to remove some of it.

This will help people access the vignettes

If a Licor data Excel file hasn't had its formulas "calculated," then many of its data columns will be read as all zeros. Eventually this will probably cause a problem with a later calculation, but it can be tough to diagnose. I could check for this when reading a file and give the user a message.

Many of the help page titles are simply the name of the function being documented, which isn't very helpful.

Maybe add an excel boolean input to read_licor_file whose default value is TRUE. If it is set to FALSE, read a plain-text log file.

Another possibility: a function that converts a plain-text file to an Excel spreadsheet. (Does this already exist?)

Many arguments to gm functions have default values. These should be removed, and the values should be specified in the analysis script. There also may be some hardcoded values that should be inputs.

When rows are deleted in an Excel file, sometimes empty rows appear at the bottom of the file. These cause problems during analysis. There should be a warning when loading a file if any blank rows are found, since it usually indicates that a mistake was made while modifying the file.

This would explain the steps in analyzing any data: translating (from the original format into R structures), validating (removing any faulty measurements), processing (e.g. applying fits to response curves), and finally analyzing (computing averages across events or species, removing statistical outliers).

The repository should probably be called PhotoGEA

This code is getting more complex, and it would be helpful to include the central functions as a library instead of having a bunch of R scripts that source each other

Steve L has pointed out that non-random residuals may indicate that an A-Ci curve was measured too slow, and vcmax actually may have changed as CO2 levels are changed.

Add a function for converting the units of a data file into coherent SI units. This will simplify a lot of calculations by removing tons of 10^n conversion factors.

Find some way to enforce rules about "data file" lists and provide some useful operations for them. This will simplify a lot of the code.

As an input argument, it would take an exdf object and a list. Each element of the list should be a list that specifies one or more points to remove. For example:

list(
  list(species = 'tobacco', plot = 5, replicate = 1, seq_num = 3),
  list(species = 'soybean', plot = c(2, 3), seq_num = 4)
)

Then all rows in the exdf object that fulfill the conditions in the specifiers should be removed.

There's probably a way to automatically identify the preamble and header rows so the user doesn't need to specify them

This function could also check that the object is a list with the proper elements, that the required elements are data frames, and that they all have the same column names. If any of these issues are detected, a warning could be issued.

• The colors supplied as the cols argument of xyplot_avg_rc get reversed. This probably shouldn't be the case.
• The X mean and standard error get calculated even when x_error_bars is FALSE. This can cause problems when the X vector is a factor rather than numeric.

I think some parts of the package could be simplified / improved with these two functions

dfoptim is used in fit_c3_aci, which appears to be a core function of the package, but dfoptim is only in the suggested packaged.