Package Structure

Structure

This is an overall structure of the samrat package, which details how the different analysis elements (parameters, analysis, plotting etc) are functionally related.

The package structure is shown below:

Samrat Structure

Directory concepts

For each analysis, we will try to keep to a one directory per analysis structure, which will include:

• Input parameter files (xlsx, csv, rds)
• Survey dataset directory

In this directory, the output of analysis will be stored in output.

This setup can then be linked with suitable reproducibility pipelines later on (cf orderly).

Conversion from mortality_small_area_estimation repo

In the originating repository the use of global variables was quite extensive and made tracking where variables originated from hard, with many variables being assigned using get and assign calls by reference to data.frames that listed the names of the variables. This made it quite hard to identify where the variable originated from and when errors appear make debugging almost impossible. In response, we choose to create names lists of variables and data sources. These are bundled within a list, in which the data.frames used to create these named lists are also included with the same name as in the original repo.

For example, previously we had a predictors xlsx file. This had a predictors_table sheet that described each of the data sets that were being imported from the other sheets as well as a dictionary sheet with more information. From this, the predictors_table sheet would be read in as predictors in the form of a data.frame. From this the actual predictor data sets would be read in and assigned to variable names determined from predictors.

Instead, we now opt for the predictors data.frame and the dictionary data.frame to be named list elements within predictors_list - this is a list object that also has a predictors_list named element that is a named list of data.frames for the actual predictor data sets. This way when referring to these, we know where they came from.