这么高深

时差=相位谱（弧度）*周期长度（月）/（2*pi）

程序运行应该是没有错误的，只是从你的quarter数据集中的变量名来看，应该是先对序列取对数然后再差分是吧？这样处理数据很难消除掉季节性因素和不规则性因素，建议先做季节性调整，去掉季节性因素和不规则因素，剩下的趋势循环因素再进行谱分析。谱分析要求样本平稳，且最好先用线性回归去除趋势，否则在谱密度估计中低频部分数值可能很高（因为含有有趋势）。

我写了一个程序，数据从硬盘导入，导入前已经经过了季节性调整（用eviews做季节性调整，tramo/seats方法），iad_r_tc表示工业增加值同比增幅经季节调整后的tc部分，_r表示同比增幅，iad_r_c_hp表示iad_r_tc再用hp滤波去除趋势。tvrs代表社会消费品零售总额。time_diff_c和time_diff_tc分别代表两序列的c部分和tc部分的时间差。对k_01_03和k_02_04做了开方，因为开方之后才是相干谱。

proc import out=work.iad
datafile="C:\Documents and Settings\EvolveGoOn\My Documents\My SAS Files\V8\my economic cycle analysis\iad\iad.xls"
DBMS=excel2000 replace;
getnames=yes;
run;
proc import out=work.tvrs
datafile="C:\Documents and Settings\EvolveGoOn\My Documents\My SAS Files\V8\my economic cycle analysis\tvrs\tvrs.xls"
DBMS=excel2000 replace;
getnames=yes;
run;
symbol i=join v=star;
proc gplot data=iad;
plot iad_r_c_hp*date;
plot iad_r_tc*date;
run;
proc gplot data=tvrs;
plot tvrs_r_c_hp*date;
plot tvrs_r_tc*date;
run;
data tvrs_iad;
set iad;
set tvrs;
run;
proc spectra data=tvrs_iad out=temp_spec_tvrs_iad a p s cross k ph adjmean whitetest;
var iad_r_c_hp iad_r_tc tvrs_r_c_hp tvrs_r_tc;
weights tukey;
run;
data spec_tvrs_iad;
set temp_spec_tvrs_iad;
time_diff_c=ph_01_03*period/(2*3.1415926535897932384626433832795);
time_diff_tc=ph_02_04*period/(2*3.1415926535897932384626433832795);
sqrt_k_01_03=sqrt(k_01_03);
sqrt_k_02_04=sqrt(k_02_04);
run;
proc export data=spec_tvrs_iad
outfile="C:\Documents and Settings\EvolveGoOn\My Documents\My SAS Files\V8\my economic cycle analysis\tvrs_iad\cspec_tvrs_iad.xls"
DBMS=excel2000 replace;
run;
proc print data=spec_tvrs_iad;
run;
proc gplot data=spec_tvrs_iad;
where period>20;
plot s_01*freq;
plot s_01*period;
plot s_02*freq;
plot s_02*period;
plot s_03*freq;
plot s_03*period;
plot s_04*freq;
plot s_04*period;
plot sqrt_k_01_03*freq;
plot sqrt_k_01_03*period;
plot sqrt_k_02_04*freq;
plot sqrt_k_02_04*period;
plot ph_01_03*freq;
plot ph_01_03*period;
plot ph_02_04*freq;
plot ph_02_04*period;
run;