No worries. Thanks. ... View more

Thanks in advance Koen, Well, my son is learning Latin at school. Hopefully that will help. ... View more

Many thanks Koen,   Unfortunately, after some digging I still am missing a final step: collecting and collating the parameter estimates and convergence statistics.   I did some research but each answer comes with three more questions. I am writing Greek without understanding a single word in Greek. The closest I get is with the following modification to your code: I managed to generate parameter estimates with names indexed by the two by groups. But the collating steps misses two features: 1. The bygroup variable cannot be populated from the PE dataset. 2. The length of PE dataset is having varying length. Never the correct length (6 = 2grpby1*3grpby2). /*Mod: turning off ODS output to not crash*/ ods graphics off; ods exclude all; ods noresults; /*Koen's main code modified lines are marked as 'mod'*/ filename deltaul clear; filename deltaul temp; data _NULL_; set work.delta_upper_limit; file deltaul; PUT 'title1 "A Transfer Function Model for the Gas Furnace Data";'; PUT 'title2 "Gas Furnace Data (grpby1 group_1 =' grpby1 +(-1) ')";'; /*Mod: I think you mean 'grpby1' here*/ PUT 'title3 "Gas Furnace Data (grpby2 group_2 =' grpby2 +(-1) ')";'; PUT 'ods output ConvergenceStatus = Convergence' grpby1 +(-1)grpby2 +(-1) ';'; /*Mod: collecting convergence statistics from each by group*/ PUT 'ods output NamedParameterEstimates = pe' grpby1 +(-1)grpby2 +(-1) ';'; /*Mod: collecting Parameter estimates from each by group*/ PUT 'proc ssm data=Seriesj_exp;'; PUT ' where grpby1="' grpby1 +(-1) '";'; PUT ' where ALSO grpby2="' grpby2 +(-1) '";'; PUT ' id obsIndex;'; PUT ' parms delta /lower=-0.9999 upper=' delta_upper_limit ';'; PUT ' state tfstate(1) T(g)=(delta) W(g)=(x3 x4 x5) ' initial_condition_state ' checkbreak;'; PUT ' comp tfinput = tfstate[1];'; PUT ' trend ar2(arma(p=2)) ;'; PUT ' intercept = 1;'; PUT ' model y = intercept tfinput ar2 ;'; PUT ' eval modelCurve = intercept + tfinput;'; PUT ' forecast out=For;'; PUT 'run;'; /*Mod: attempting to collect parameter estimates (PE) from the double by group and add a flag (bygroup)*/ PUT 'data pe;'; PUT 'set pe pe' grpby1 +(-1)grpby2 +(-1) ';'; PUT 'bygroup = by' grpby1 +(-1)grpby2 +(-1) ';'; PUT 'run;'; run; title; footnote; %INCLUDE deltaul / source2; title; footnote; filename deltaul clear; /*Mod: turning ODS output back on*/ ods graphics on; ods exclude none; ods results; /* end of program */ I have added a 'mod' to wherever I made the modfiication.   Would you be able to nudge it into place?   Thanks again.   Hao ... View more

No worries, please take your time. Appreciate it greatly. ... View more

Hi Koen, Is there any examples? Thanks. ... View more

Yes, please. That would greatly help. Additionally, similar to setting the parameter limit per by group, I need to set the initial state per by group. If possible, some notes on where the macro is loading up the limits/initial states, doing the looping and collecting the estimates would be great. ... View more

Thanks Koen, I think I can vaguely infer what is being done here. The sticking point for me is my complete lack of understanding on Macros. May I ask a few follow up questions: 1. I rely on the ODS output function to collect the fitted estimates such as the state time series and parameter estimates. The HTML output crashes always with that many by groups. I wonder whether/where those were incorporate in your code? 2. The by group seems to be your 'grpby' variable. Does that mean I need to flatten my double by group (stock and day) into just one? ... View more