Description Usage Arguments Details Value See Also Examples
View source: R/coin_audit_tools.R
This is an analysis function for seeing what happens when elements of the composite indicator are removed. This can help with "what if" experiments and acts as different measure of the influence of each indicator or aggregate.
1  removeElements(COIN, aglev, isel, quietly = FALSE)

COIN 
The COIN, which must be constructed up to and including the aggregation step. 
aglev 
The level at which to remove elements. For example, 
isel 
A character string indicating the indicator or aggregate code to extract from each iteration. I.e. normally this would be set to
the index code to compare the ranks of the index upon removing each indicator or aggregate. But it can be any code that is present in

quietly 
Logical: if FALSE (default) will output to the console an indication of progress. Might be useful when iterating over many indicators. Otherwise set to TRUE to shut this up. 
One way of looking at indicator "importance" in a composite indicator is via correlations. A different way is to see what happens if we remove the indicator completely from the framework. If removing an indicator or a whole aggregation of indicators results in very little rank change, it is one indication that perhaps it is not necessary to include it. Emphasis on one: there may be many other things to take into account.
This function works by successively setting the weight of each indicator or aggregate to zero. If the analysis is performed at the indicator level, it creates a copy of the COIN, sets the weight of the first indicator to zero, regenerates the results, and compares to the nominal results (results when no weights are set to zero). It repeats this for each indicator in turn, such that each time one indicator is set to zero weights, and the others retain their original weights. The output is a series of tables comparing scores and ranks (see Value).
Note that "removing the indicator" here means more precisely "setting its weight to zero". In most cases the first implies the second, but check that the aggregation method that you are using satisfies this relationship. For example, if the aggregation method does not use any weights, then setting the weight to zero will have no effect.
A list with elements as follows:
.$Scores
: a data frame where each column is the scores for each unit, with indicator/aggregate corresponding to the column name removed.
E.g. .$Scores$Ind1
gives the scores resulting from removing "Ind1".
.$Ranks
: as above but ranks
.$RankDiffs
: as above but difference between nominal rank and rank on removing each indicator/aggregate
.$RankAbsDiffs
: as above but absolute rank differences
.$MeanAbsDiffs
: as above, but the mean of each column. So it is the mean (over units) absolute rank change resulting from removing each
indicator or aggregate.
compTable()
Comparison table between two COINs
compTableMulti()
Comparison table between multiple COINs
1 2 3 4 5 6 7 8  # check the effect of removing ASEM subpillars, one at a time
# First build ASEM index
ASEM < build_ASEM()
# now run check at subindex level (level 3), on index scores/ranks
CheckPillars < removeElements(ASEM, 3, "Index")
# summary by pillar
CheckPillars$MeanAbsDiff
# have a look at the rest of the output list to see more details.

Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.