ABSTRACT

Suppose now that we have two different and independclit variables, XI and X2 which are iiormally distributed with nlcails O and varia,nccs l , and subject t,o independent rneasurerncnt errors Yl illid YA with variancc n? If WC want to devote two bits of storage space to tlic resulting observations Zl and Z2 , it would be sensible to uscl the signs of XI and x2.