Date: Wed, 5 May 1999 07:16:29 +0200 z@cati.si "SPSSX(r) Discussion" Zenel Batagelj Weights <19990505040036.119EF1F001@planja.arnes.si> text/plain; charset="iso-8859-2"

Actually I think weights are not the best solution for your problem, whay don't you use the Agregate function (Data menu)? You just aggregate your date according to account id and what you do you simoply calculate average sccores for each of the agregate...

Otherwise I think you did some kind of mistake... check your syntax again.

zenel

> > I have a weighting problem and I need to know a way to adjust spss so it > > calculates the weights according to a client's specifications. > > There are four accounts in a line of business. We carried out customer > > satisfaction surveys in each of these accounts. There were a different > > number of respondents in each account. > > Acct 1: 4 > > Acct 2: 5 > > Acct 3: 38 > > Acct 4: 74 > > > > This line of business was compared with another line of business, which > > had 30 accounts, but only one respondent per account. > > > > I calculated a weight where > > wt = 1 if it was in the LOB of 30 accounts. > > wt = 1/4 if it was in the first account of the first LOB > > wt = 1/5 if it was in the second account... > > wt = 1/38 for the third > > wt = 1/74 for the fourth > > > > and used these weights to calculate means of the satisfaction scores for > > each of the accounts. > > > > If there are an inadequate number of respondents due to missings in the > > weighted accounts, then spss assesses the account and decides > there isn't > > enough weight to calculate a mean. That means in some cases there are > > only 3 accounts represented in the first LOB instead of 4. The client > > doesn't like that. > > > > Is there a way to tell spss to readjust the calculation to > accept as many > > valid respondents as there are and adjust the weight accordingly? > > > > thanks!

Back to: Top of message | Previous page | Main SPSSX-L page