COVID19データを分析する-週別
前回に引き続きCOVID19データを分析します。
今回は取り込んだCOVID19データの週の何曜日から連続して感染者が発生しているかカウントします。
こうしてみると火曜日、木曜日、土曜日始まりのケースが目立ちますね。
サラリーマンの通勤行動に何かしら因果関係があるのでしょうか。
前回の連続感染者の増加がみられたケースの集計結果をstart_dateのweekdayのカテゴリで集計してみた結果になります。
プログラム
filename resp "/home/xxxxxxx/excel/covid.txt";
proc http url='https://covid.ourworldindata.org/data/owid-covid-data.json'
method="GET" out=resp;
run;
/* Assign a JSON library to the HTTP response */
libname space JSON fileref=resp;
*assign country;
%let cntry=JPN;
data have;
drop ordinal:;
retain year month day _date dif_new_cases;
set space.&cntry._data;
by date;
format _date yymmdds10.;
dif_new_cases=dif(new_cases);
_date=input(date, yymmdd10.);
year=year(_date);
month=month(_date);
day=day(_date);
output;
drop date;
rename _date=date;
run;
data out;
length year month day max_dif_new_cases max_dif_new_cases_date max_new_cases
max_new_cases_date max_new_deaths max_new_deaths_date max_stringency_index
max_stringency_index_date 8;
format date max_dif_new_cases_date max_new_cases_date max_new_deaths_date
max_stringency_index_date yymmdds10.;
do until(last.month);
set have end=lr curobs=_r;
by month notsorted;
if max_dif_new_cases < dif_new_cases then
do;
max_dif_new_cases=dif_new_cases;
max_dif_new_cases_date=date;
end;
if max_new_cases < new_cases then
do;
max_new_cases=new_cases;
max_new_cases_date=date;
end;
if max_new_deaths < new_deaths then
do;
max_new_deaths=new_deaths;
max_new_deaths_date=date;
end;
if max_stringency_index < stringency_index then
do;
max_stringency_index=stringency_index;
max_stringency_index_date=date;
end;
end;
drop new_cases-numeric-new_tests_per_thousand tests_units;
run;
data want;
length year month day dif_new_cases max_dif_new_cases max_dif_new_cases_date
new_cases max_new_cases max_new_cases_date new_deaths max_new_deaths
max_new_deaths_date stringency_index max_stringency_index
max_stringency_index_date 8;
format date max_dif_new_cases_date max_new_cases_date max_new_deaths_date
max_stringency_index_date yymmdds10.;
do until(last.month);
set have;
by month notsorted;
dif_new_cases=ifn(first.month, ., dif_new_cases);
if max_dif_new_cases < dif_new_cases then
do;
max_dif_new_cases=dif_new_cases;
max_dif_new_cases_date=date;
end;
if max_new_cases < new_cases then
do;
max_new_cases=new_cases;
max_new_cases_date=date;
end;
if max_new_deaths < new_deaths then
do;
max_new_deaths=new_deaths;
max_new_deaths_date=date;
end;
if max_stringency_index < stringency_index then
do;
max_stringency_index=stringency_index;
max_stringency_index_date=date;
end;
end;
do until(last.month);
set have;
by month notsorted;
output;
end;
drop new_cases-numeric-new_tests_per_thousand tests_units;
run;
options fullstimer locale=en_us;
*連続感染者の推移をランキング出力;
data consecevt(keep=con_cases exact_count total_count)
consecevt_detail(keep=start_date end_date con_cases weekday)
weekday_cnt(keep=weekday con_cases start_date weekday_cnt);
if _N_=1 then do;
if 0 then set want;
dcl hash consecevt(ordered:"d",suminc:"exact_count",multidata:"y")
consecevt_det(ordered:"d",suminc:"exact_count",multidata:"y")
out1(multidata:"y",ordered:"d")
out2(multidata:"n",ordered:"d")
out3(multidata:"y",ordered:"d");
consecevt.definekey("con_cases");
consecevt.definedata("con_cases","year","month","day");
consecevt.definedone();
consecevt_det.definekey("con_cases");
consecevt_det.definedata("con_cases","year","month","day");
consecevt_det.definedone();
out1.definekey("con_cases");
out1.definedata("con_cases","exact_count","total_count");
out1.definedone();
out2.definekey("start_date");
out2.definedata("start_date","end_date","con_cases");
out2.definedone();
out3.definekey("weekday");
out3.definedata("weekday","con_cases","start_date");
out3.definedone();
dcl hiter ci("consecevt_det") out2i("out2") out3i("out3");
end;
format start_date end_date yymmdds10.;
con_cases=0;
do until(last.month);
set want end=lr;
by year month notsorted;
con_cases=ifn(sign(dif_new_cases)=1,con_cases+1,0);
if sign(dif_new_cases)=1 then consecevt.ref();
if con_cases then consecevt_det.add();
end;
if lr;
retain exact_count 1;
total_adjust=0;
*感染者推移ランキング;
do con_cases=consecevt.num_items to 1 by -1;
consecevt.sum(sum:exact_count);
total_count=exact_count+total_adjust;
output consecevt;
total_adjust+exact_count;
end;
*詳細;
do while(ci.next()=0);
end_date=mdy(month,day,year);
start_date=end_date-con_cases+1;
_iorc_=out2.add();
end;
if lr;
do while(out2i.next()=0);
weekday=weekday(start_date);
_iorc_=out3.add();
output consecevt_detail;
end;
if lr;
do while(out3i.next()=0);
weekday_cnt=ifn(lag(weekday) ne weekday,1,sum(weekday_cnt,1));
output weekday_cnt;
end;
run;
*曜日表示用出力形式;
proc format;
picture mywk(default=10)
1="日"
2="月"
3="火"
4="水"
5="木"
6="金"
7="土"
;
run;
*上のコードに次期に修正する予定;
data weekday_cnt(keep=weekday weekday_cnt);
do until(last.weekday);
format weekday mywk.;
set weekday_cnt;
by weekday notsorted;
end;
run;
*csvファイル作成マクロ;
%macro callcsv(filename,dsn);
ods csv file="&filename";
proc print data=&dsn noobs;
title "consecevt";
footnote "&sysdate";
run;
ods csv close;
%mend;
*csvファイル作成バッチ;
data _null_;
infile datalines dlm="09"x;
input filename:$200. dsn:$50.;
*_iorc_=dosubl(cats('%callcsv(',filename,',',dsn,')'));
call execute(cats('%callcsv(',filename,',',dsn,')'));
datalines;
/home/xxxxxxx/excel/consecevt.csv consecevt
/home/xxxxxxx/excel/consecevt_detail.csv consecevt_detail
/home/xxxxxxx/excel/weekday_cnt.csv weekday_cnt
;
run;
