var numericColumnStats = function( data, keys ){
    var console = console || { debug: function(){} },
        stats = {};

    // precondition: keys is an array of keys accessible on each data in data
    if( !keys || !keys.length ){
        throw new Error( 'keys is a required parameter and must be an array:' + keys );
    }
    if( !data || !data.length ){
        return stats;
    }

    function parseVal( val ){
        if( val === null ){
            throw new Error( 'Null value' );
        }
        val = Number( val );
        if( isNaN( val ) || !isFinite( val ) ){
            throw new Error( 'NaN or non-finite number' );
        }
        return val;
    }

    // does this effect the data in the other thread?
    [ 'min', 'max', 'sum', 'mean', 'median', 'count' ].forEach( function( name, i ){
        stats[ name ] = new Array( keys.length );
    });

    var keyIndex = 0,
        separatedCols = new Array( keys.length );
    keys.forEach( function( key, keyIndex ){
        stats.min[ keyIndex ] = 0;
        stats.max[ keyIndex ] = 0;
        stats.sum[ keyIndex ] = 0;
        separatedCols[ keyIndex ] = [];
    });

    // work backwards to prevent co-modification problems while splitting up data into columns
    for( var dataIndex=( data.length - 1 ); dataIndex>=0; dataIndex-=1 ){
        var datum = data.pop( dataIndex );

        for( keyIndex=0; keyIndex<keys.length; keyIndex+=1 ){
            var key = keys[ keyIndex ],
                datumColVal = datum[ key ];

            try {
                //NOTE: parsing value as number
                datumColVal = parseVal( datumColVal );
            } catch( e ){
                continue;
            }

            // separate the columns
            separatedCols[ keyIndex ].unshift( datumColVal );
            // get the other stats
            stats.min[ keyIndex ] = Math.min( stats.min[ keyIndex ], datumColVal );
            stats.max[ keyIndex ] = Math.max( stats.max[ keyIndex ], datumColVal );
            stats.sum[ keyIndex ] += datumColVal;
        }
    }

    // get counts, mean, median
    function comparator( a, b ){
        if( a < b ){ return -1; }
        if( a < b ){ return 1; }
        return 0;
    }

    for( keyIndex=0; keyIndex<keys.length; keyIndex+=1 ){
        var count = separatedCols[ keyIndex ].length,
            sum = stats.sum[ keyIndex ];
        stats.count[ keyIndex ] = count;
        stats.mean[ keyIndex ] = ( sum / count );

        // sort columns for median
        separatedCols[ keyIndex ].sort( comparator );

        // odd count -> straight forward median
        var middleDataIndex = Math.floor( count / 2 );
        if( count % 2 === 1 ){
            stats.median[ keyIndex ] = separatedCols[ keyIndex ][ middleDataIndex ];
        } else {
            var middleValA = separatedCols[ keyIndex ][ middleDataIndex ],
                middleValB = separatedCols[ keyIndex ][( middleDataIndex + 1 )];
            stats.median[ keyIndex ] = ( middleValA + middleValB ) / 2;
        }
    }
    return stats;
};