The document details a currency volatility project conducted in SAS. It involves importing currency exchange rate data from text files, merging the datasets, calculating log returns and volatility, and analyzing the correlation and distributions of the returns through various SAS procedures and statistical tests. Temporary and permanent SAS datasets are created to store the cleaned, merged and transformed data for further analysis.

1. /*CURRENCY VOLATILITY PROJECT
Author: Phil Duhe, November, 2009
***************************PART A******************************************/
*Create two temporary data sets that contain the original data in the
text files Data1.txt and Data2.txt;
*system options for page appearance;
*Establish a file reference for data1.txt;
filename DATA1 'C:NewSASDATA1.TXT';
*reading the data from data1.txt;
data A;
infile DATA1 firstobs=7;
input date yymmdd6. dow bp cd dm;
*Change all the 0 values to missing;
if bp <= 0 then bp = .;
if cd <= 0 then cd = .;
if dm <= 0 then dm = .;
run;
*establish file reference for data2.txt;
filename DATA2 'C:NewSASDATA2.TXT';
*reading the data from data2.txt;
data B;
infile DATA2 firstobs=7;
input date: yymmdd6. dow jy sf;
*Change all the 0 values to missing;
if jy <= 0 then jy = .;
if sf <= 0 then sf = .;
run;
************************END OF PART A*********************************;
***************************PART B*************************************;
*Sorting data for merge;
proc sort data=A;
by date dow;
run;
proc sort data=B;
by date dow;
run;
*************************PART B C D &E**********************************;
*Creating permanent merged data set MyLib.MergedData;
options ls=80 ps=60 pageno=1 nodate center;
libname MyLib 'C:SASIntro';
data MyLib.MergedData;
merge A(in=ina) B(in=inb);
by date dow;
if ina and inb;*merged set contains nonmissing observations in A and B;
format date mmddyy8. dow 2.0 bp--sf 7.6;
*Labeling the variables(PART E);

2. label date = 'Date'
dow = 'Day of Week'
bp = 'British Pound'
cd = 'Canadian Dollar'
dm = 'Deutsche Mark'
jy = 'Japanese Yen'
sf = 'Swiss Franc';
run;
/*Print output*/
proc print data=MyLib.MergedData Width=FULL label;
title 'Merged Data Set';
*show date and not #observations;
id date;
var dow bp--sf;
run;
******************************END OF PART B D
&E****************************;
***********************************PART
C*************************************;
data MyLib.LogResult (drop=bp cd dm jy sf
zbp zcd zdm zjy zsf);
set MyLib.MergedData;
zbp = bp / (lag(bp)) ;
zcd = cd / (lag(cd)) ;
zdm = dm / (lag(dm)) ;
zjy = jy / (lag(jy)) ;
zsf = sf / (lag(sf)) ;
*Calculate the log of daily returns(PART C);
lgbp = log(zbp);
lgcd = log(zcd);
lgdm = log(zdm);
lgjy = log(zjy);
lgsf = log(zsf);
*Format and label the variables;
format date mmddyy8. dow 2.0 lgbp--lgsf 7.6 ;
*Labeling the variables(PART E);
label date='Date'
dow ='Day of Week'
lgbp='British Pound'
lgcd='Canadian Dollar'
lgdm='German Deutschmark'
lgjy='Japanese Yen'
lgsf='Swiss Franc';
run;
options ls=80 ps=60 pageno=1 nodate center;
proc print data=MyLib.LogResult Width=Full label;
title 'Log of Daily Returns';
id date;
var dow lgbp--lgsf;
run;
*Volatility PART C;
/*Generate standard deviations of random variables

3. create a new data set based on PROC MEANS*/
proc means data=MyLib.LogResult std n noprint;*no need to print means
statistics;
var lgbp--lgsf;
*Outputting the Results into the dataset VolResult;
output out=MyLib.VolResult(drop= _TYPE_ _FREQ_)
std=S_bp S_cd S_dm S_jy S_sf
n = N_bp N_cd N_dm N_jy N_sf;
run;
/*Print to verify standard deviations*/
proc print data=MyLib.VolResult Width=Full label;
options ls=80 ps=60 pageno=1 nodate center;
title 'Standard Deviations of Log of Daily Returns';
var S_bp--S_sf;
run;
*Calculate the estimate of annual volatility;
data MyLib.AnnVol(drop=S_bp S_cd S_dm S_jy S_sf
N_bp N_cd N_dm N_jy N_sf);
set MyLib.VolResult;
*Annualized volatility calculation;
Vol_bp =S_bp*sqrt(250/N_bp);
Vol_cd =S_cd*sqrt(250/N_cd);
Vol_dm =S_dm*sqrt(250/N_dm);
Vol_jy =S_jy*sqrt(250/N_jy);
Vol_sf =S_sf*sqrt(250/N_sf);
label Vol_bp = 'British Pound'
Vol_cd = 'Canadian Dollar'
Vol_dm = 'German Deutschmark'
Vol_jy = 'Japanese Yen'
Vol_sf = 'Swiss Franc';
run;
proc print data=MyLib.AnnVol Width= FULL label;
title1 'Annualized Foreign Exchange Rate Volatilty';
title2;
title3 'Major Trading Currencies';
run;
********************************END OF PART C******************************;
*******************************PART F**************************************;
/*Create an external file in (ASCII format) using PUT statement
and appropriate FORMAT statement*/
proc sort data=MyLib.LogResult;
by date dow;
run;
Data _null_;
set MyLib.LogResult;
file 'C:UsersPhilSasProject'; **write merged data to output
file;
if _n_=1 then do;*do loop to establish report headings and column spacings;
* report headings;
put 'SAS X405, UCBX, November 2009, Term Project';
put 'Phil Duhe';

4. put ' ';
put 'Natural Logarithm of Daily Returns';
put 'of Five Currencies';
put 'June, 1973 to November, 1987';
put ' ';
put ' ';
put ' ';
put ' '
' British'
' Deutsche'
' Canadian'
' Japanese'
' Swiss'
;
put ' '
' Pounds'
' Marks'
' Dollars'
' Yen'
' Franc'
;
put ' ';
put ' Date'
' LGBP'
' LGCD'
' LGDM'
' LGJY'
' LGSF'
;
put ' - ';
end; * end if 1st obs;
put @2 date MMDDYY8. @15lgbp 7.6 @25lgcd 7.6 @35lgdm 7.6 @45lgjy
7.6 @55lgsf 7.6;
run;
******************************END OF PART
F**********************************;
********************************PART
G***************************************;
/*Calculate the variance-covariance matrix of the daily return of random
variables*/
/*Note: The variance of a variable is a measure of the dispersion of the
variable
around its mean value.*/
/*Note: Covariance is a measure of the strength of the link between two
random numerical
variables*/
ods select Cov;
proc corr data=MyLib.LogResult cov nocorr nosimple;*just print the variance-
covariance matrix;
options ls=180 ps=65 pageno=1 nodate center;*landscape;
title1 'Variance-Covariance Matrix';
title2;
title3 'Log of Daily Returns of Major FX Currencies';

5. var lgbp--lgsf;
run;
**********************************END OF PART G*****************************;
************************************PART
H***********************************;
/*calculate the variance-covariance matrix for Mondays*/
data MyLib.Monday;
set MyLib.LogResult;
where dow EQ 1;*read input data set for Monday only;
run;
ods select Cov;
proc corr data=MyLib.Monday cov nocorr nosimple ;*print covariances and not
descriptive statistics;
options ls=180 ps=60 pageno=1 nodate nocenter;*landscape;
title 'Monday Trading Days';
var lgbp--lgsf;
run;
/*calculate the variance-covariance matrix for Thursdays*/
data MyLib.Thursday;
set MyLib.LogResult;
where dow EQ 4;*read input data set for Thursday only;
run;
ods select Cov;
proc corr data=MyLib.Thursday cov nocorr nosimple;*print covariances and not
descriptive statistics;
options ls=180 ps=60 pageno=1 nodate nocenter;*landscape;
title 'Thursday Trading Days';
var lgbp--lgsf;
run;
***********************************END OF PART H**************************;
**************************************PART I******************************;
/*Calculate the correlation coefficient matrix of the log of daily returns*/
proc corr data=MyLib.LogResult Pearson nosimple nomiss rank;
options ls=180 ps=60 pageno=1 nodate nocenter;*landscape;
title1 'Correlation Coefficient Matrix';
title2;
title3 'Log of Daily Returns of Major FX Currencies';
var lgbp--lgsf;
run;
***********************************END OF PART I***************************;
**************************************PART J*******************************;
/*Calculate the skewness, kurtosis of the log of daily return of all the
random variables*/
/*Skewness: measures the tendency of the deviations to be larger in one
direction than in the other. The
sample sknewness can be either positive or negative. Observations that have
a normal distribution
should have a sknewness near zero(0).*/

6. /*Kurtosis: measures the heaviness of tails of the data distribution.
Observations that are normally
distributed should have a kurtosis near zero(0).*/
proc means data=MyLib.LogResult skewness kurtosis;
options ls=80 ps=60 pageno=1 nodate center;
title 'Skewness and Kurtosis of Log of Daily Returns';
var lgbp--lgsf;
run;
**********************************END OF PART J*************************;
****************OPTIONAL************************************************;
/*EXERCISE K*/
/*Write a macro program so that you can repeatedly excecute correlation
coefficients for different data sets without retyping the procedure*/
%macro example(request= , mydata= );*parameter variables set to null;
%if &request=print %then
%do;
proc print data=&mydata;*print data set;
run;
%end;
%else %if &request=correlation %then*run the correlation;
%do;
proc corr data=&mydata pearson nosimple nomiss rank;*rank the
correlations;
run;
%end;
%mend example;
***********************************************************************;
*Example of Calling MACRO Example;
%example(request=correlation, mydata=MyLib.LogResult)*pass in parameter
variables;
***********************************************************************;
/*Extra Programs*/
/*Univariate Statistics for Log of Daily Returns*/
ods select BasicMeasures ExtremeObs;*Basic statistics and extreme
observations for
nominal value of fx currencies;
title "Basic Measures";
proc univariate data=MyLib.LogResult;
var lgbp--lgsf;
run;
title "Log of Daily Returns vs. Normal Distribution";
proc univariate data=MyLib.LogResult noprint;
var lgbp--lgsf;
histogram lgbp--lgsf/normal (color=red w=2) ;
inset mean='Mean' (7.6)
std ='Standard Deviation' (7.6)/font='Arial'
pos=nw
height=3;

7. run;
*******************************************************************;
ods html;
ods graphics on;
proc corr data=MyLib.LogResult nomiss noprint plots=matrix;
var lgbp--lgsf;
title 'Correlation Coefficients Using PROC CORR';
run;
ods graphics off;
ods html close;
********************************************************************;
ods html;
ods graphics on;
proc corr data=MyLib.LogResult nomiss noprint
plots=scatter(alpha = .20 .30);
var lgbp--lgsf;
run;
ods graphics off;
ods html close;