Example workflow

If one has a csv that contains the data, the first question is whether the frame contains columns that are needed for the indexing. Should this be the case one can use the readAndIndexFrame, is that not the case one has first to reading the frame with readFrame add the columns that are needed for indexing and then index with indexWithColumnValues. You will need to index with at least 2 columns should you wish to aggregate because then can have the same Key value pairs in one column as needed/wished for the aggregation. Each time you want to aggregate means you have to have an additional column for the indexing. Keep in mind that you can use functions of the GroupTransform module as often as you want because the row keys are not changed. In this example we go with 2 aggregations because we have biological and technical replicates.

let frameForTutorial = 
  frame [
    ("ConditionA")=> series ["row1" => 2.;"row2" =>3.;"row3" =>1.;"row4" =>7.]
    ("ConditionB")=> series ["row1" => 2.4;"row2" =>4.5;"row3" =>6.1;"row4" =>5.1] 
  ]
  |>Frame.addCol "Gen"(series ["row1" => "A" ;"row2" =>"A";"row3" =>"A";"row4" =>"A"])
  |>Frame.addCol "technicalReplicate"(series ["row1" => "B" ;"row2" =>"B";"row3" =>"C";"row4" =>"C"])
  |>Frame.addCol "BioRep"(series ["row1" => "D" ;"row2" =>"E";"row3" =>"D";"row4" =>"E"])



let indexedTutorialFrame = indexWithColumnValues ["Gen";"technicalReplicate";"BioRep"] frameForTutorial

indexedTutorialFrame.Print()

ConditionA ConditionB Gen technicalReplicate BioRep 
Gen: A
technicalReplicate: B
BioRep: D
 -> 2          2.4        A   B                  D      
Gen: A
technicalReplicate: B
BioRep: E
 -> 3          4.5        A   B                  E      
Gen: A
technicalReplicate: C
BioRep: D
 -> 1          6.1        A   C                  D      
Gen: A
technicalReplicate: C
BioRep: E
 -> 7          5.1        A   C                  E

!!!! Keep in mind that the function working on the whole frame will error should you have strings as values inside the frame for the NumericAggregation function or ints/floats for the StringAggregation. So remove the values/columns or use the single column versions multiple times. !!!!

let indexedFrameWithoutIndexingColumns = 
  indexedTutorialFrame
  |>Frame.dropCol "Gen"
  |>Frame.dropCol "technicalReplicate"
  |>Frame.dropCol "BioRep"

indexedFrameWithoutIndexingColumns.Print()

ConditionA ConditionB 
Gen: A
technicalReplicate: B
BioRep: D
 -> 2          2.4        
Gen: A
technicalReplicate: B
BioRep: E
 -> 3          4.5        
Gen: A
technicalReplicate: C
BioRep: D
 -> 1          6.1        
Gen: A
technicalReplicate: C
BioRep: E
 -> 7          5.1

As you can see each row has multiple Key objects and the combinations are unique. Now one can create a filter with either the NumericFilter or GroupFilter module functions. to use the resultant series or seq series one has to keep in mind that the functions in the NumericAggregation and StringAggregation modules need for each column a seq of series, so one needs to create a seq of seq of series for the function that affect the whole frame. If not done correctly this could result in an error(an empty filter is not suitable for the functions working on the whole frame!). When you have a suitable seq of series or seq of seq of series then one can then use the aggregation module of choice. Let's look back at our example with filters that filters would always say true, we want the Mean, and we use the seq ["Gen";"technicalReplicate"] to determine that Rep.

let filterA = numericFilterServeralCol (IsBiggerThan 0.5) indexedFrameWithoutIndexingColumns
let filterB = numericFilterServeralCol (IsSmallerThan 100.) indexedFrameWithoutIndexingColumns

let seqOfFilterMulty filterOne filterTwo = 
  filterOne
  |>Seq.mapi (fun i x -> 
    let colKeysForSeq seqIt b= 
      seqIt
      |>Seq.item b
    let cIII = seq [x;colKeysForSeq filterTwo i]
    cIII)

let aggregatedFrameA = numAgAllCol Mean indexedFrameWithoutIndexingColumns ["Gen";"technicalReplicate"] (seqOfFilterMulty filterA filterB)
aggregatedFrameA.Print()

ConditionA ConditionB 
Gen: A
technicalReplicate: B
 -> 2.5        3.45       
Gen: A
technicalReplicate: C
 -> 4          5.6

As you can see the BioRep parts of the keys was dropped and we aggregated Keys that were identical. Now we also want to do the median of the technicalReplicate. For that we need new filters and need to adjust the input parameters.

let filterC = numericFilterServeralCol (IsBiggerThan 0.5) aggregatedFrameA 
let filterD = numericFilterServeralCol (IsSmallerThan 100.) aggregatedFrameA 

let aggregatedFrameB = numAgAllCol Median aggregatedFrameA ["Gen"] (seqOfFilterMulty filterC filterD)

aggregatedFrameB.Print()

ConditionA ConditionB 
Gen: A
 -> 3.25       4.525