TEST SOLUTIONS, Stat 430  In-Class Test S09

ANSWERS: 
 
I. (a)    PROC FREQ data=home.times; 
               tables AMT * GRP /  
                 norow nocolumn nocum nopercent ; 
 
   (b)    PROC CORR data=home.times; 
               var liftim; with amt; 
        /* for partial correlation add the line */ 
               partial grp; 
 
   (c)    PROC MEANS DATA=HOME.TIMES min max; 
               var LIFTIM; 
               class AMT GRP ; 
               output out=homout min=LIFMIN max=LIFMAX; 
          data homrang(keep = AMT GRP RANGE); 
               set homout; 
               RANGE = LIFMAX-LIFMIN; 
               if AMT ne . and GRP ne . ; 
          proc print data=homrang; run; 


This whole problem part could also have been done using the OUTPUT
statement from PROC UNIVARIATE. For OUTPUT unlike the printed output,
you can specify what statistics will be calculated, and RANGE is an
allowed keyword for that output.

As it turns out, in current version of SAS "range" is also an allowed
statistic keyword for PROC MEANS, so that (using range in  place of
min max) is definitely the best way to do problem I(c).

              
    (d)   PROC UNIVARIATE data=home.times ; 
               histogram liftim; 
               class grp; run; 
        or 
 
          PROC GCHART data=home.times; 
              vbar liftim / group = grp ; run; 
 
 
   In either case, you wanted to specify as an option:
           LEVELS=5   ,      or
           midpoints a to b by c

II. (a) Needed to plot step function with upward jumps of 
1/8 at each observation except for 7, whre the jump is 1/4.
Find quantile as midpoint of interval of values for which the
empirical distribution function crosses or is equal to the 
fraction (1/4 for Q1, 1.2 for median) desired.

> quantile(tdat,c(.25,.5), type=2) 
25% 50%  
  6   7  
 
(b). > hist(tdat, breaks=seq(4,14,2), plot=F)[c(1,2,4,5)] 
$breaks 
[1]  4  6  8 10 12 14 
$counts 
[1] 2 3 0 2 1 
$density 
[1] 0.1250000 0.1875000 0.0000000 0.1250000 0.0625000 
$mids 
[1]  5  7  9 11 13 
 
NOTE observation falling on interval boundary

 
(III.) Discussed in class. Words matter !