options ls=80 nocenter nodate;

* You should run this program and bring the output to class. *;
* I will be referring to the output by page number.          *;

* Data on weight, tar content, nicotine content, and
carbon monoxide (CO) content for 25 brands of cigarettes;

data cigdat;
input id $ brand $ tar nico weight co;
cards;
A Alpine          14.1  .86  .9853  13.6
B Benson&Hedges   16.0 1.06 1.0938  16.6
C BullDurham      29.8 2.03 1.1650  23.5
D CamelLights      8.0  .67  .9280  10.2
E Carlton          4.1  .40  .9462   5.4
F Chesterfield    15.0 1.04  .8885  15.0
G GoldenLights     8.8  .76 1.0267   9.0
H Kent            12.4  .95  .9225  12.3
I Kool            16.6 1.12  .9372  16.3
J L&M             14.9 1.02  .8858  15.4
K LarkLights      13.7 1.01  .9643  13.0
L Marlboro        15.1  .90  .9316  14.4
M Merit            7.8  .57  .9705  10.0
N MultiFilter     11.4  .78 1.1240  10.2
O NewportLights    9.0  .74  .8517   9.5
P Now              1.0  .13  .7851   1.5
Q OldGold         17.0 1.26  .9186  18.5
R PallMallLight   12.8 1.08 1.0395  12.6
S Raleigh         15.8  .96  .9573  17.5
T SalemUltra       4.5  .42  .9106   4.9
U Tareyton        14.5 1.01 1.0070  15.9
V True             7.3  .61  .9806   8.5
W ViceroyRichLight 8.6  .69  .9693  10.6
X VirginiaSlims   15.2 1.02  .9496  13.9
Y WinstonLights   12.0 .82  1.1184  14.9
;

* Sort the data by nico and then print out the dataon nico and co. *;
proc sort; by nico;
proc print; var id brand nico co;

* Graph of co vs nico. *;
proc plot;
plot co*nico=id;
run;

* Regression of co vs nico.                                           *;
* I will explain the options r and influence in class.                *;
* The ODS command puts certain output results into a new SAS dataset. *;

proc reg;
model co = nico / r influence;
ods output OutputStatistics=outstat1;
run;

* Univariate statistics on some of the REG output. I will explain this in class. *; 
proc univariate plot normal;
var cooksd dfb_intercept dfb_nico dffits hatdiagonal studentresidual;
run;

* We change the CO value for data point C by 5 units and rerun the analysis *;

data newdat1; set cigdat; if (id = 'C') then co=co-5; 
proc plot; plot co*nico=id; run;
proc reg; model co = nico / r influence; 
ods output OutputStatistics=outstat2; run;
proc univariate plot normal;
var cooksd dfb_intercept dfb_nico dffits hatdiagonal studentresidual; run;

* We now change the CO value for data point C back to its original value, and instead  *;
* change the value of data point L by 5 units. We then rerun the analysis again.       *;

data newdat2; set cigdat; if (id = 'L') then co=co-5; 
proc plot; plot co*nico=id; run;
proc reg; model co = nico / r influence; ods output OutputStatistics=outstat2; run;
proc univariate plot normal;
var cooksd dfb_intercept dfb_nico dffits hatdiagonal studentresidual; run;

* Generate some additional output that I will explain in class. *;
proc iml;
use cigdat var {nico};
read all;
n = nrow(nico);
x=j(n,1)||nico;
a = x*inv(t(x)*x);
create dfzstat from a [colname={'dfbz_int' 'dfbz_nico'}];
append from a;
proc print data=zstat;
proc univariate plot; var dfbz_int dfbz_nico;
run;