Notes to contributors

Stick to the structure, design choices and style conventions described below. For questions: please contact the author.

General design choices

  • The only OO system used is the S3 system. No S4 classes will be allowed!
  • Whenever you subset a matrix using [ use: [..., ..., drop = FALSE] to avoid accidentally dropping the dim attributes.
  • Generally avoid using attributes (sometimes they are meaningful though)
  • Whenever the output consists of more than 1 element use a named list!

Style/Naming

Stick to this styleguide with the following exceptions/additions:

  1. Function and class names are always CamelCase. Function names should contain a verb followed by a noun like: processData(), calculateValue().
  2. Verbs in function names should be consistent across the whole package. Avoid to mix synonyms. Example: computeValue() vs. calculateValue(). This package always uses calculate instead of compute. Similarly, method vs e.g. approach. This package always uses approach instead of method.
  3. Use plural in function/object names if the main output is more than one element, like scaleWeights(), calculateComposites(), handleArgs() etc. but stick to singular in other cases like parseModel().
  4. Strive for meaningful argument names even if they are a bit longer than usual. People are much better at remembering arguments like respect_structural_model compared to something like resp_sm. Naming should also be consistent if possible. For example: any argument that describes a method or an approach should be named .appraoch_*.
  5. Argument names always start with a dot to distinguish them from other objects.
  6. Indentation: It is OK to align function arguments indented by two spaces below the function name instead of where the function starts if this help with clarity.
## Both ok but second is prefered in this case
calculateInnerWeightsPLS <- function(.S                           = NULL,
                                     .W                           = NULL,
                                     .csem_model                  = NULL,
                                     .PLS_weight_scheme_inner     = c("centroid",
                                                                      "factorial", 
                                                                      "path"),
                                     .PLS_ignore_structural_model = NULL
                                     ) { }

calculateInnerWeightsPLS <- function(
  .S                            = NULL,
  .W                            = NULL,
  .csem_model                   = NULL,
  .PLS_weight_scheme_inner      = c("centroid","factorial", "path"),
  .PLS_ignore_structural_model  = NULL
  ) { }

Matrices

  • Whenever possible: variables belong in columns, observations belong to rows. This implies: whenever a matrix contains observations, variables are in columns, no matter if they are indicators, proxies, errors or anything else.
  • Covariance matrices: indicators always belong to columns and proxies to rows, i.e., the matrix of weights W for PLS is therefore (J × K) where J is the number of proxies and K the number of indicators.
  • Naming: matrices within the package should be named according to the naming schemes of the related SEM literature.

Arguments

All arguments used by any of the functions in the package (including internal functions) are centrally collected in the file zz_arguments.R. Whenever a new argument is introduced:

  1. Add the new argument name + a description to the cSEMArguments list in alphabetical order by writing @param <argument> "Some description". "Defaults to xxx".
  2. Add the argument to the args or the args_dotdotdot_csem list of the args_default() function and provide a default value.
    • Arguments for args_dotdotdot_csem: all arguments that can be used when calling csem() or cca() . Practically this comprises all formal arguments of foreman() that are not formal arguments of csem().
    • Arguments for args: all other arguments.
  3. Add the argument to the function you want to use it in. Order arguments according to their importance (i.e. .data and .model always come first). if there is one otherwise use alphabetical order.