/*This macro is designed to remove the data for the probes       */
/*that are effectively off.  Probes that are effectively off     */
/*are probes that have a signal that is less than the 95th        */
/*percentile of the signal of the negative controls per slide    */
/*per dye at least 50% of the time for all treatments.           */

%macro find_p95(lib,infile,probe_id,factor1,factor2,signal,
neg_control,neg_con_ind,trt);
*reassigned infile to work folder;
	data &infile._1;
	set &infile;
	run;

*took p90 for each slide/dye combo;
	proc means data=&infile._1 p95 noprint;
		/*remove comments if variable denoting negative controls*/
		/*is a numeric*/
		where &neg_control =&neg_con_ind;
		/*remove comments if variable denoting negative controls*/
		/*is a character*/
		/*where &neg_control ="&neg_con_ind";*/
		class &factor1 &factor2;
		var &signal;
		output out=&factor1&factor2._p95 p95=p95;
	run;

*subsetted _type_=1;
	data &factor1&factor2._p95_1;
	set &factor1&factor2._p95;
		where _type_=3;
	run;

*sorted by slide/dye;
	proc sort data=&infile._1;
		by &factor1 &factor2;
	run;

*merged p95 for each slide/dye combo to original file;
	data p95_merge;
	merge &factor1&factor2._p95_1 &infile._1;
		by &factor1 &factor2;
	run;

*assigned spot on/off for each spot;
	data spot_off;
	set p95_merge;
		if &signal ge p95 then spot_off=0;
		if &signal<p95 or iswellabovebg=0 then spot_off=1;
	run;

*sorted by &probe_id;
	proc sort data=spot_off;
		by &probe_id;
	run;

*calculated means of each spot by &probe_id;
	proc means data=spot_off noprint;
		class &trt;
		var spot_off;
		by &probe_id;
		output out=outoff mean=mean;
	run;

*subsetted _type_=1;
	data outoff1;
	set outoff;
		where _type_=1;
	run;

*saved to permenant data set;
	data outoff1;
	set outoff1;
	run;

%mend find_p95;

%macro per_off (trt,trtn);
*percent off per trt determined by the mean of the ?;
	data po_&trt;
	set outoff1;
		where &trt="&trtn";
		percent_off_&trtn=mean;
		if percent_off_&trtn>.5 then off_&trtn=1;
			else off_&trtn=0;
		keep &probe_id percent_off_&trtn off_&trtn;
	run;
%mend per_off;

%macro find_off(lib,infile,probe_id,factor1,factor2,signal,
neg_control,neg_con_ind,trt,trt1,trt2,trt3,trt4,trt5,trt6,
trt7,trt8,trt9,trt10,suffix);
*find p95;
	%find_p95(&lib,&infile,&probe_id,&factor1,&factor2,
	&signal,&neg_control,&neg_con_ind,&trt);

*find the percent off for each trt;
	%per_off(&trt,&trt1);
	%per_off(&trt,&trt2);
	%per_off(&trt,&trt3);
	%per_off(&trt,&trt4);
	%per_off(&trt,&trt5);
	%per_off(&trt,&trt6);
	%per_off(&trt,&trt7);
	%per_off(&trt,&trt8);
	%per_off(&trt,&trt9);
	%per_off(&trt,&trt10);

*merge all the percent off data sets together;
	data partial_off&suffix;
	merge po_&trt1 po_&trt2 po_&trt3 po_&trt4
		po_&trt5 po_&trt6 po_&trt7 po_&trt8
		po_&trt9 po_&trt10;
		by &probe_id;
	run;

*gene is off if it is off at every trt;
	data gene_off;
	set partial_off&suffix;
		gene_off=0;
		if off_&trt1=1 and off_&trt2=1 
			and off_&trt3=1 and off_&trt4=1
			and off_&trt5=1 and off_&trt6=1 
			and off_&trt7=1 and off_&trt8=1
			and off_&trt9=1 and off_&trt10=1 
			then gene_off=1;
	run;

	proc sort data=gene_off;
		by &probe_id;
	run;

*create data set with only off genes;
	data off_list_&lib._&suffix;
	set gene_off;
		where gene_off=1;
		if first.&probe_id;
		by &probe_id;
	run;

*sort orig file by probeuid;
	proc sort data=&infile;
		by &probe_id;
	run;

*merge orig file and gene off/on calls;
	data anova_all;
	merge &infile gene_off;
		by &probe_id;
	run;

*subset all on genes into permenant data set;
	data anova_nooff_&lib._&suffix;
	set anova_all;
		where gene_off ne 1;
	run;
%mend find_off;