Without the time series your data will work as is but you'll have to change the variable names below to what you end up with. The column headings for the monthly values aren't valid SAS variable namea and you may either end up with _08_31_2012 or VAR1 by default depending on import method. In some respects if they come in as VAR1 through VAR12 the coding may be easier. /* trend = (Average of 12 th Month Balance - Average of 1 st month Balance)/Average of 1 st Month Balance*/ trend = (VAR12 - VAR1)/ VAR1; 2. /*Short_term_growth= (Maximum Balance over last 3 months - Latest Balance)/latest balance */ Short_term_growth = (max( of var10-var12) - Var12)/ Var12; 3 /*Long_term_growth= (Maximum Balance over 12 Months – Latest Balance)/Latest Balance */ long_term_growth = (max( of var1-var12) - Var12)/ Var12; growth = VAR12 - mean(of var1-var12); And if you sort by client ID you could reshape the data as recommended by ETS_KPS above and run a simple linear regression on model of value = date and look at the slope parameter and statistics which might be telling if there is a big change in balance between the last and next-to-last month. Consider what happens if months 2 to 11 show and average of 100 increase in balance but the month 12 balance is roughly the same as the month 1 balance.
... View more