proc format;
value WEEK
1="2/20" 2="2/27" 3="3/06" 4="3/13"
5="3/20" 6="3/27" 7="4/03" 8="4/10"
9="4/17" 10="5/08" 11="5/15" 12="5/22";
run;

/* Coupons in the Sunday paper, such as
   SmartSource (TM) Magazine, RedPlum.com (TM), and P&G BrandSaver. 
   I record three kinds of coupons:
   1) General coupons that can be redeemed at any grocery store and are valid for a certain value
      such as $0.50 off, $1.00 off, or $5.00 off. These are entered by value, followed by the 
      number of coupons of that value.
   2) Coupons that are "buy one, get one free" (BOGO), up to a certain price.
      The maximum value is recorded as a single entry with the comment "BOGO."
   3) Store-specific coupons such as for Red Lobster or Boston Market.
      These are lumped together into a single entry with the comment "Store."
   */
/* general coupons, by value and count */
data Coupons;
infile datalines missover;
input Week Value Freq Comment $;
format Week WEEK. Value 5.2;
datalines;
1 0.25 1 
1 0.50 3
1 1.00 13
1 1.50 4
1 2.00 3
1 3.00 3
1 5.00 1
2 0.25 4
2 0.35 1
2 0.5 15
2 0.55 4
2 0.75 10
2 1.00 47
2 1.5  4
2 2.00 11
2 2.50 2
2 3.00 7
2 3.50 1
2 4.00 5
2 5.00 6
2 10.00 2
3 0.25 1
3 0.30 1
3 0.35 1
3 0.40 7
3 0.50 8
3 0.55 3
3 0.75 15
3 1.00 50
3 1.50 11
3 2.00 11
3 3.00 5
3 4.00 2
3 6.00 1
3 10.00 1
4  0.25 1
4  0.3  1
4  0.35 1
4  0.4  3
4  0.5  13
4  0.55 2
4  0.6  1
4  0.75 12
4  1.00 40
4  2.00 5
4  3.00 7
4  3.5  1
4  5.00 2
4 10.00 1
5 0.35 1
5 0.4  1
5 0.5  13
5 0.75 3
5 1.00 50
5 1.25 1
5 1.5  8
5 2.00 8
5 2.5  2
5 3.00 6
5 4.00 3
5 5.00 1
5 7.00 1
6 0.35 1
6 0.4  3
6 0.5  16
6 0.55 3
6 0.6  1
6 0.7  1
6 0.75 14
6 1.00 45
6 1.5  4
6 2.00 10
6 3.00 4
6 4.00 2
6 5.00 3
6 7.00 1
6 10.00 1
7 0.3  1
7 0.35 3
7 0.4  1
7 0.5  21
7 0.55 4
7 0.75 12
7 1.00 57
7 1.5  5
7 2.00 13
7 3.00 3
7 4.00 3
7 5.00 4
7 7.00 1
7 10.00 1
7 15.00 1
8 0.25 5
8 0.35 2
8 0.4  2
8 0.5  9
8 0.55 4
8 0.6  3
8 0.75 14
8 1.00 44
8 1.25 1
8 1.5  7
8 2.00 17
8 2.5  1
8 3.00 12
8 4.00 1
8 5.00 5
8 6.00 1
8 10.00 1
9 0.25 2
9 0.4  6
9 0.5  19
9 0.55 9
9 0.6  3
9 0.7  1
9 0.75 18
9 1.00 78
9 1.25 3
9 1.50 15
9 2.00 25
9 2.50 3
9 3.00 14
9 4.00 5
9 5.00 3
9 6.00 1
9 7.00 1
10 0.25 6
10 0.35 5
10 0.4  1
10 0.45 2
10 0.5  27
10 0.55 8
10 0.6  1
10 0.65 1
10 0.75 12
10 1.00 79
10 1.25 2
10 1.5  12
10 2.00 22
10 2.50 2
10 3.00 12
10 4.00 7
10 5.00 10
10 10.00 2
10 15.00 1
11 0.25 1
11 0.3  2
11 0.35 1
11 0.5  12
11 0.55 3
11 0.6  1
11 0.75 7
11 1.00 59
11 1.25 1
11 1.5  10
11 2.00 20
11 2.50 1
11 3.00 8
11 4.00 2
11 5.00 4
11 10.00 1
12  0.25 3
12  0.3  1
12  0.35 3
12  0.4  3
12  0.5  20
12  0.55 6
12  0.6  1
12  0.65 1
12  0.75 16
12  1.00 60
12  1.25 1
12  1.50 7
12  2.00 15
12  3.00 3
12  4.00 2
12 10.00 1
;
run;

ods graphics on;
title "Distribution of Coupon Amounts";
proc freq data=Coupons;
weight freq;
table Value /plots=freqplot(scale=percent);
run;

proc freq data=Coupons order=freq;
weight freq;
table Value /plots=freqplot(scale=percent);
run;

proc means data=Coupons N Sum Mean noprint;
by Week;
freq Freq;
var Value;
output out=General N=N Sum=General;
run;

/* Store-specific and BOGO coupons. 
   Also the newspaper's CLAIM of maximum savings */
data Others;
input Week Store BOGO Claim;
format Week WEEK. Store 5.2 BOGO 5.2 Claim 5.2;
datalines;
 1  5.00  2.88  43
 2  0.00 59.42 242
 3 19.00 17.39 150
 4 15.00 38.17 149
 5 16.00  9.48 153
 6 10.00 18.21 164
 7 27.50 22.20 210
 8  0.00 24.96 227
 9 17.00 16.97 208
10 29.00 41.73 331
11  7.00  6.87 207
12 17.99 23.97 181
;
run;


data Results;
merge General Others;
Total = General+Store+BOGO;
GpB = General+BOGO;
Diff = Total - Claim;
label Total = "Actual Potential Savings"
      Claim = "Claimed Savings"
      Diff = "Difference between Actual and Claimed Values";
run;

title "Maximum Savings By Week";
proc print;
var Week General BOGO Store Total Claim;
run;

data Results;
set Results;
x=331; y=331; /* add maximum values for identity line */
run;

title "Predicted versus Actual Maximum Savings";
proc sgplot data=Results noautolegend;
/* identity line */
vector x=x y=y /xorigin=43 yorigin=43 noarrowheads;
scatter x=Claim y=Total;
xaxis grid;
yaxis grid;
run;

title "Difference between Actual and Claimed Values";

proc univariate data=Results;
var DIff;
histogram Diff / grid midpoints=0 to 130 by 20 vscale=count;
run;


proc sort data=results out=f; by Diff;run;
proc freq data=f order=data; tables Diff; run;

title;
proc robustreg data=Results method=lts;
model Total = Claim;
run;

proc reg data=Results;
model Total = Claim;
run;