/*get rid of commitment to work variable*/
/*specifying library*/
%let direc=C:\AAA PROJECTS\ECBW\N2P;
libname here "&direc" ;
libname there "&direc.data7" ;
option spool;

data finaldata;set here.finaldata;
if F1weight > -9;

/*Just last minute variable recoding*/
drop commit ;
if race=2 then race2="Hispanic";
if race=1 then race2="Asian";
if race=3 then race2="Black";
if race=4 then race2="White";
if race=5 then race2="Native";

/*creating TS_talkA*/
/*Oct 2002, following Charle's hunch, I am using the original coding here*/
if TS_talk2=2 then TS_talkA="2S talk two Ts";
if TS_talk2=1 then TS_talkA="1S talk one T";
if TS_talk2=0 then TS_talkA="3S talk zero T";
*if TS_talk2 < -999 then TS_talkA="4missing";

closureX=closure;

if closure = 0 /*or closure=.5*/ then CLOSE=0;
if closure=.5 or closure = 1 or closure=1.5   then CLOSE=1;
if closure=2 or closure =2.5  or closure=3  or closure=3.5 or closure=4  then CLOSE=2;
if CLOSE=0 then CLOSE0=1;
if CLOSE=1 then CLOSE1=1;
if CLOSE=2 then CLOSE2=1;

/*risk factor*/
if BYRISK >=0 and BYRISK < = 6 then do;
RISK0=0;
RISK1=0;
end;

if BYRISK=0 then RISK0=1;
if BYRISK> 0 and BYRISK =<6 then RISK1=1;

/*delition flag due to missing cases*/
/*tedeous because these were stirng variables*/
%macro nan (var1=,var2=);
miss&var2=0;
if &var1  = "" then miss&var2=1;
%mend nan;
%nan (var2=1,var1=gender);
%nan (var2=2,var1=race2);
%nan (var2=3,var1=urban);
%nan (var2=4,var1=F1schooltype);
%nan (var2=5,var1=F1palvalue);
%nan (var2=6,var1=F1clubtime);
*%nan (var2=7,var1=F1goodteacher);/*I took this out.  This came from F2 cycle*/
%nan (var2=8,var1=F1getalong);
%nan (var2=9,var1=F1tutor);
%nan (var2=10,var1=TS_talkA);
%nan (var2=11,var1=F1famact);

X1=sum(of miss1 miss2 miss3 miss4 miss5 miss6 /*miss7*/ miss8 miss9 miss10 miss11);
drop miss1 miss2 miss3 miss4 miss5 miss6 /*miss7*/ miss8 miss9 miss10 miss11;
/*easy because these are numerics*/
X2=nmiss(of F1compo_score f1ses F1G10enrol  parentschool closure /* commitR*/);
X3=X1+X2;
valid=0;if X3=0 then valid=1;

/*just student teacher tie information missing*/
nodata=0;
if TS_talk2 < -999 then nodata=1;

/*data missing*/
if valid =1 then nodata2=0;
if valid =0 then nodata2=1;

whole="whole";

/*creating 1 0 dummy variables for descriptive statistics, so I can use proc means instead of proc freq*/
if race > -9 then do;
hispanic=0;asian=0;black=0;white=0;native=0;
if race=2 then hispanic=1;
if race=1 then asian=1;
if race=3 then black=1;
if race=4 then white=1;
if race=5 then native=1;
end;
misrace=0;
if race < -9  then misrace=1;

urban2=0;rural=0;suburb=0;
if urban="(Urb)" then urban2=1;
if urban="(Rur)" then rural=1;
if urban="(Sub)" then suburb=1;

if BYgetalong="S not get along with T" then BYgetalong2=0;
if BYgetalong="S get along with T" then BYgetalong2=1;

if F1getalong="S not get along with T" then F1getalong2=0;
if F1getalong="S get along with T" then F1getalong2=1;

if TS_talkA ne "" then do;
talk0=0;talk1=0;talk2=0;
if TS_talkA="3S talk zero T" then talk0=1;
if TS_talkA="1S talk one T" then talk1=1;
if TS_talkA="2S talk two Ts" then talk2=1;
end;

private=0;catholic=0;public=0;
if F1schooltype="1Private" then private=1;
if F1schooltype="2Catholic" then catholic=1;
if F1schooltype="3Public" then public=1;

If F1famact="2FamTalk No" then F1famact2=0;
If F1famact="1FamTalk Yes" then F1famact2=1;

If F1palvalue="2FamTalk No" then F1palvalue2=0;
If F1palvalue="1FamTalk Yes" then F1palvalue2=1;
run;



data finaldata;set finaldata;
if X=1 then closurerank2="highclose";
if X=0 then closurerank2="lowclose";
if valid=0 then valid2="not valid";
if valid=1 then valid2="valid";

changeCompo_score=F1compo_score-Bycompo_score;

run;


%let vars=
F1problem BYproblem  
BF1devRlog
F1college BYcollege 

asian black hispanic native
F1ses F1compo_score 

rural suburb urban2

F1G10enrol
private catholic 

F1famact2
F1palvalue2 
F1clubtime 
parentschool 

F1getalong2
F1tutor
talk0 talk1 talk2
closure


late
cut
skip
cheat
copy
fight
gang
racist
sexist
steal
destroy
smoke
drink
drug
weapon
abuse
talkback
disobey;



/*First I get descriptive statistics of the NELS data, the whole
thing, for the variables used in my analysis.  After this,
I get the same statistics for my analytical sample, the data
that I actually used.  I wanted to compare the whole sample
with the smaller data that I used.  The data became small
because of missing cases.  The rate of missing was very severe,
so I wanted to compare the two samples*/
proc means  data=finaldata mean std min max n stderr;
weight F1weight;
class gender;
weight F1weight;
var &vars;
ods output summary=whole;
run;

data whole;set whole;
length ID $ 10;
x="NELS";
ID=x||gender;
drop gender;
run;

/*GET descriptive statistics for subgroups within data*/
/*In this case, boys and girls' samples*/
/*Note the variable "valid2" that I used at where-line*/
/*In this way I pick cases that have no missing values*/
/*See how valid2 is constructed above*/
%macro jones (var1=, and=);
proc means  data=finaldata  mean std min max n stderr;
weight F1weight;
where valid2="valid";
class &var1;
weight F1weight;
var 
&vars
;
ods output summary=&var1;
run;

data &var1;set &var1;
length ID $ 10;
ID=&var1;
drop &var1;
run;
%mend jones;
*%jones (var1=valid2);
%jones (var1=gender);
*%jones (var1=race2);
*%jones (var1=urban);
*%jones (var1=closurerank2);
*%jones (var1=ts_talkA);
*%jones (var1=F1schooltype);

/*Merging the two result data sets*/
data allthese;
set whole gender;
run;

/*Transpose to get the data into a shape*/
proc transpose data=allthese out=alltheseT;
id ID;
run;


/*Lots of manipulation to put the data into a shape
that is ready to go into an excel sheet*/
data alltheseT2;set alltheseT;
x=_name_;
group=tranwrd(x,"_Mean","");
group=tranwrd(group,"_StdDev","");
group=tranwrd(group,"_Min","");
group=tranwrd(group,"_N","");
group=tranwrd(group,"_Max","");
group=tranwrd(group,"_StdErr","");
run;

/*I do proc print once in a while to see how the data looks like*/
proc print;
run;

proc print data=alltheseT;
where _label_="Mean" or _label_="Std Dev" or _label_="N Obs";
run;


/*Sometimes I have to do a complex way of PROC TRANSPOSE*/
/*I don't enjoy this sort of programming*/
%macro john (var1=);
proc transpose data=alltheseT2 out=&var1;
by group notsorted;
id _label_;
var  &var1;
run;

data &var1;set &var1;
keep group
&var1.mean
&var1.std_
&var1.max 
&var1.min 
&var1.stdev
&var1.N
&var1.std_error
;

if minimum=0 and maximum=1 then do;
x=(mean*(1-mean));
Std_Error=sqrt(x/(N) );
end;
drop x;

&var1.mean=mean;
&var1.std_error=Std_Error;
&var1.max =maximum;
&var1.min = minimum;
&var1.stdev=std_dev;
&var1.N=N;
run;


%mend john;
%john (var1=_BOY_);
%john (var1=_GIRL_);
%john (var1=NELS_BOY_);
%john (var1=NELS_GIRL_);


data descriptivestat;
/*retain statement changes the order of variables in the data set*/
/*In this case I did not need it, so I did not use it*/
*retain NELS_BOY_mean	NELS_BOY_std_error	NELS_BOY_max	NELS_BOY_min	NELS_BOY_stdev	NELS_BOY_N	
_BOY_mean	_BOY_std_error	_BOY_max	_BOY_min	_BOY_stdev	_BOY_N	
NELS_GIRL_mean	NELS_GIRL_std_error	NELS_GIRL_max	NELS_GIRL_min	NELS_GIRL_stdev	NELS_GIRL_N
_GIRL_mean	_GIRL_std_error	_GIRL_max	_GIRL_min	_GIRL_stdev	_GIRL_N	
;
merge NELS_BOY_  _BOY_   NELS_GIRL_ _GIRL_;
*by group;
run;

proc print;run;

PROC EXPORT DATA= work.descriptivestat
            OUTFILE= "C:\Documents and Settings\kuekawa\My Documents\tables.xls" 
            DBMS=EXCEL2000 REPLACE;
RUN;