appropriate degrees of freedom, 9 in this case (N-1). CDF produces this with the inverse probability option:

CDFCDF (T, DF=9, INVERSE) .05 T05 ;

T05 is a scalar variable that is returned with the argument of the t-distribution at the 5 percent level. Note that we do not need to get the 95 percent level separately because the t is a symmetric distribution centered around a known constant (zero).

T05 is the value of the t-statistic for a standardized variable. To obtain the confidence interval bounds in units corresponding to the logarithm of rainfall, we use the equation given in P&R (eq. 2.15):

s_x is the standard deviation of the log of rainfall, which was also printed in the MSD table we obtained in section 5. Here is the completed program for computation of the confidence interval:

SET LRLO = @MEAN-T05*@STDDEV/SQRT(@NOB) ;
SET LRHI = @MEAN+T05*@STDDEV/SQRT(@NOB) ;
PRINT @MEAN LRLO LRHI ;

Note how we used @NOB, the number of observations in the current sample for N. Variables beginning with @ are special variables stored automatically by TSP.

To get the confidence interval in terms of the original rainfall units, we use some more SET statements:

SET RLO = EXP(LRLO) ;
SET RHI = EXP(LRHI) ;
SET RMEAN = EXP(@MEAN) ;