/*This program demonstrates how to work with serial data using transpose, arrays, do loops and dim function; 
Also does some plotting.  Data are borrowed from Professor Lardaro's RI economic indicators web site
http://members.cox.net/lardaro/current.htm*/

libname lab 'C:\1datalap\learning\psy532x'; run;

PROC IMPORT OUT= x 
            DATAFILE= "C:\1dataLap\personal\econ_ind.xls" 
            DBMS=EXCEL REPLACE;
     
     GETNAMES=NO;
     MIXED=NO;
     SCANTEXT=YES;
     USEDATE=YES;
     SCANTIME=YES;
RUN;

proc print; run;
/*flip the data*/
proc transpose data=x out=vert;
run;

proc print data = vert; run;

data x1; set x; 
/*Make one vertical column of all data--'stacking'*/
array econ(1:12) F2-F13;
   do k=1 to 12;
          if econ(k)= 58 then econ(k)=.;
		  econ1 = econ{k}; 
		  output;
		  end;		  
		drop  F2-F13;
run;

proc print; run;

data lab.serial_ex; set x1;
time = _N_;
if time le 12 then year = '1983';
if time ge 13 and time lt 27 then year = '1984';
/*etc.*/
run;

proc print; var time k year ; run;
/*now try some descriptives and plotting*/

proc capability data = lab.serial_ex; var econ1; histogram econ1; run;

proc gplot data=lab.serial_ex;
      symbol i=spline v=dot h=2;
      plot econ1 * time;
run;

/*what the heck, try some time series (assuming the economy has no seasonality, does not need to be differenced, and follows
AR(1)--in other words, rather unthoughtfully*/

proc arima data=lab.serial_ex;
	identify var = econ1;
	estimate p=1 q=1;
run;

/*Assignment option 1: go to http://www.technion.ac.il/docs/sas/ets/chap2/sect27.htm and try some other time series plots
due in two weeks*/

data missing_ex1; set x;
/*set up a flag variable for missingness*/
miss_flag1=0;
If F2 in (50,83) then miss_flag1 = 1; 
miss_flag2=0;
If F3 in (50,83) then miss_flag2 = 1; 
miss_flag3=0;
If F4 in (50,83) then miss_flag3 = 1; 
Where_miss = (100 * miss_flag3) + (10 * miss_flag2) + (miss_flag1);
run;

proc print data = missing_ex1; var Where_miss miss_flag1 miss_flag2 miss_flag3; run;

Proc freq data = missing_ex1;
tables Where_miss;
*By group; /*Assignment option 2: create a group and then compare missingness for two groups*/
run;

data missing_ex2; set x; 
/*making one vertical column of all data, again*/
array econ(1:12) F2-F13;
   do k=1 to 12;
          if econ(k) in (58,83,67) then econ(k)=.;
		  econ1 = econ{k}; 
		  output;
		  end;		  
		drop k;
run;
data missing_ex2a; set missing_ex2; 
time = _N_;
run;

proc print data = missing_ex2a; run;
proc gplot data=missing_ex2a;
      symbol i=spline v=circle h=2;
      plot econ1 * time;
run;

proc arima data=missing_ex2a;
	identify var = econ1;
	estimate p=1 q=1;
run;

/*using arrays to recode all the variables for given criteria*/

data missing_ex3; set x;
/*recode all values of 58 ACROSS VARIABLES to missing*/
array econ(1:12) F2-F13;
   do k=1 to 12;
          if econ(k)= 58 then econ(k)=.;
		  econ1 = econ{k}; output;
		  end;
		drop k;
run;
quit;
Proc means n nmiss data=missing_ex3; run;

proc print data = missing_ex3; var econ1; run;
data missing_ex4; set x;
/*recode all the variables to missing, for all numeric variables*/
Breed = 'doberman';
if F2 = 100 then breed = '  ';
array nvar(*) _numeric_;
     do i=1 to dim(nvar);  
          if nvar(i)=67 then nvar(i)=.;
     end;
run;
quit;
proc print data = missing_ex4; *var F13; run;
Proc means n nmiss data=missing_ex4; run;

/*Assignment option 3: use the above code and other tricks that you have learned to calculate the overall
percent missingness in the database*/

/*Assignment option 4: create some additional time series type plots using any of the above databases; optionally:
experiment with adding economic indicators and reviewing correlations or cross correlations in the TS framework;  
Lardaro's web page has several options;

*/

Data x1; set x;
January = F2;
February = F3;
March = F4;
April = F5;
May = F6;
June = F7;
July = F8;
August = F9;
September = F10;
October = F11;
November = F12;
December = F13;
YEAR = F1;
drop F1-F13;
run;