Question

I’m trying to get a correlation matrix in SQL Server and my data is in a table the following way:

RptLOB1     RptLOB2   Correlation
AE          AE             1
Bail        AE            0.35
Commercial  Bail          0.25
Commercial  AE            0.15

…and so on.

I want to write a code so my output looks the following way:

            AE     Bail   Commercial
AE          1      0.35      0.15
Bail        0.35    1        0.25
Commercial  0.15   0.25       1

Order of the RptLOB doesn’t matter as long as the order is the same from top to bottom and left to right on top. I’ve been trying to find a way to approach this and I’m not quite sure what the best way is. I was thinking using PIVOT but that will not output the RptLOB’s on top (they will be considered as columns in the table).

EDIT:

This output is going to be inserted in another table like so:

col1             col2        col3                            col4        col5              

Generic
Company Inputs   Insurance   Stochastic Model Correlations   Exposure    Correlation Matrix
                 AE          Bail                            Commercial
AE               1           0.35                            0.15
Bail             0.35        1                               0.25
Commercial       0.15        0.25                            1

Answer 1

You can use a PIVOT for this. If you know the number of columns you have to transform, then you can use a static version:

select *
from 
(
  select RptLOB1 RptLOB1, RPTLOB2 RPTLOB2, Correlation
  from yourtable
  union all
  select RPTLOB2, RptLOB1, Correlation
  from yourtable
  union all
  select distinct RptLOB1, RptLOB1, 1.0
  from yourtable
) x
pivot
(
  max(Correlation)
  for RPTLOB2 in ([AE], [Bail], [Commercial])
) p;

see SQL Fiddle with demo

If you have an unknown number of values to correlate, then you will want to use a dynamic version:

DECLARE @query  AS NVARCHAR(MAX),
    @colsPivot as  NVARCHAR(MAX)

select @colsPivot = STUFF((SELECT distinct  ',' 
                      + quotename(RptLOB1)
                    from yourtable t
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')

set @query 
  = 'select RptLOB1, '+@colspivot+ '
     from 
     (
       select RptLOB1 RptLOB1, RPTLOB2 RPTLOB2, Correlation
       from yourtable
       union all
       select RPTLOB2, RptLOB1, Correlation
       from yourtable
       union all
       select distinct RptLOB1, RptLOB1, 1.0
       from yourtable
      ) x
      pivot
      (
        max(Correlation) 
        for RPTLOB2 in ('+ @colspivot +')
      ) p'

exec(@query)

see SQL Fiddle with demo

Edit – based on your comments, if you want the column headers in another row, then you can use the following:

DECLARE @query  AS NVARCHAR(MAX),
    @colsPivot as  NVARCHAR(MAX),
    @colsRow as  NVARCHAR(MAX),
    @colsConverted as  NVARCHAR(MAX)

select @colsPivot = STUFF((SELECT distinct  ',' 
                      + quotename(RptLOB1)
                    from yourtable t
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')

select @colsRow = STUFF((SELECT distinct  ', ''' 
                      + RptLOB1 + ''' as ' + RptLOB1
                    from yourtable t
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')

select @colsConverted
      = STUFF((SELECT distinct  ', CAST(' 
                + quotename(RptLOB1) 
                 + ' as varchar(50))'
                from yourtable t
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')


set @query 
  = 'select ''RptLOB1'' as RptLOB1, 
      '+ @colsRow + '
     union all 
     select RptLOB1, '+ @colsConverted+ '
     from 
     (
       select RptLOB1 RptLOB1, RPTLOB2 RPTLOB2, Correlation
       from yourtable
       union all
       select RPTLOB2, RptLOB1, Correlation
       from yourtable
       union all
       select distinct RptLOB1, RptLOB1, 1.0
       from yourtable
      ) x
      pivot
      (
        max(Correlation) 
        for RPTLOB2 in ('+ @colspivot +')
      ) p'

exec(@query)

see SQL Fiddle with demo