@Vince28: Sorry, I mispoke I suppose, you are correct, the order of the variables does not matter, by permutations I really meant including the variable or not, which is much less than 14!, but is still far too many to manually code by hand. -As for the model forms, basically written out my full model is this: "10YrTr" and "3MoTr" being U.S. treasury interest rates as my two independent variables, and superscript +/- is for a positive change verse a negative change in these variables from the previous obseration (t-1) to current (t). Time "t-1" represents a lag of one month, time "t" represents the current month. "R" is the dependent Y variable which could be a number of interest rates that I amt rying to model. Permutations include removing some of these terms that may not reflect behaviors significant for modeling a certain rate. Variations of the model would be like for example: /*ALL Variables: */ proc model data=to_forecast_parms outparms=parms_all_Asymm;   dif_&term_rate. =B1*dif_10YrTr_up + B2*dif_10YrTr_down +  B3*lag_dif_10YrTr_up + B4*lag_dif_10YrTr_down +   B5*dif_3MoTr_up + B6*dif_3MoTr_down + B7*lag_dif_3MoTr_up + B8*lag_dif_3MoTr_down + B9*lag_dif_&term_rate. +   Gamma*(A0 + A1*lag_3MoTr + A2*lag_10YrTr + A3*(lag_3MoTr*lag_3MoTr) + A4*(lag_10YrTr*lag_10YrTr) +   A5*(lag_3MoTr*lag_3MoTr*lag_3MoTr) + A6*(lag_10YrTr*lag_10YrTr*lag_10YrTr) - lag_&term_rate.);   fit dif_&term_rate. / dw=1 collin white; run; /*No Instantaneous 10YrTr: */ proc model data=to_forecast_parms outparms=parms_all_Asymm;   dif_&term_rate. = B3*lag_dif_10YrTr_up + B4*lag_dif_10YrTr_down +   B5*dif_3MoTr_up + B6*dif_3MoTr_down + B7*lag_dif_3MoTr_up + B8*lag_dif_3MoTr_down + B9*lag_dif_&term_rate. +   Gamma*(A0 + A1*lag_3MoTr + A2*lag_10YrTr + A3*(lag_3MoTr*lag_3MoTr) + A4*(lag_10YrTr*lag_10YrTr) +   A5*(lag_3MoTr*lag_3MoTr*lag_3MoTr) + A6*(lag_10YrTr*lag_10YrTr*lag_10YrTr) - lag_&term_rate.);   fit dif_&term_rate. / dw=1 collin white; run; /*No Instantaneous 3MoTr */ proc model data=to_forecast_parms outparms=parms_all_Asymm;   dif_&term_rate. = B3*lag_dif_10YrTr_up + B4*lag_dif_10YrTr_down +   B7*lag_dif_3MoTr_up + B8*lag_dif_3MoTr_down + B9*lag_dif_&term_rate. +   Gamma*(A0 + A1*lag_3MoTr + A2*lag_10YrTr + A3*(lag_3MoTr*lag_3MoTr) + A4*(lag_10YrTr*lag_10YrTr) +   A5*(lag_3MoTr*lag_3MoTr*lag_3MoTr) + A6*(lag_10YrTr*lag_10YrTr*lag_10YrTr) - lag_&term_rate.);   fit dif_&term_rate. / dw=1 collin white; run; /*No Instantaneous or lagged 10YrTr */ proc model data=to_forecast_parms outparms=parms_all_Asymm;   dif_&term_rate. = B7*lag_dif_3MoTr_up + B8*lag_dif_3MoTr_down + B9*lag_dif_&term_rate. +   Gamma*(A0 + A1*lag_3MoTr + A2*lag_10YrTr + A3*(lag_3MoTr*lag_3MoTr) + A4*(lag_10YrTr*lag_10YrTr) +   A5*(lag_3MoTr*lag_3MoTr*lag_3MoTr) + A6*(lag_10YrTr*lag_10YrTr*lag_10YrTr) - lag_&term_rate.);   fit dif_&term_rate. / dw=1 collin white; run; /*No Instantaneous or lagged 10YrTr or mean reversion */ proc model data=to_forecast_parms outparms=parms_all_Asymm;   dif_&term_rate. = B7*lag_dif_3MoTr_up + B8*lag_dif_3MoTr_down + B9*lag_dif_&term_rate.;   fit dif_&term_rate. / dw=1 collin white; run; /*No Instantaneous 3MoTr or mean reversion */ proc model data=to_forecast_parms outparms=parms_all_Asymm;   dif_&term_rate. = B3*lag_dif_10YrTr_up + B4*lag_dif_10YrTr_down +   B7*lag_dif_3MoTr_up + B8*lag_dif_3MoTr_down + B9*lag_dif_&term_rate.;   fit dif_&term_rate. / dw=1 collin white; run; ___________________________________________________________________________ These are just a couple of examples, I am just excluding some behaviors in each of the forms that may  not affect a particular rate that I am trying to model. @Reeza: " Will this info be sufficient? Model b1, b2, b3, b4, b5, b6, b7, b8, b9, gamma, a0, a1, a2, a3, a4, a5, a6, R-square" -- I would like to add Adj-R-square, RMSE, and white's value for heterscedasicity test if possible - BUT the above is the minimum requirement, and should be sufficient if it is significantly harder to add the others. I do agree with create the master dataset of all model estimates/statistics as the process goes, rather than keeping the datasets hanging around, anything I can do to optimize is beneficial. Please let me know if I can provide any more information on what I am attempting. ... View more

Sorry if I wasn't clear enough about the output, as long as I can grab the results from the outparms= dataset, and at least the R-square value and as I go through the macros (or even after the fact as long as the datasets for each model are uniquely named) append the results to a master dataset where each observation of the dataset is the parameters estimates and the r-square value i.e. the master dataset may have column headers such as: b1, b2, b3, b4, b5, b6, b7, b8, b9, gamma, a0, a1, a2, a3, a4, a5, a6, R-square with a row containing the values for each model permutation (the model forms that don't include certain coefficients will just have that coefficient with a value of '0' in the file). Does this make sense? Forgive me if I am not being clear. Thanks !! ... View more

Reeza, This is my proc model statement for the full model form with all variables (some variables don't need to be permuted i.e. B9, Gamma, A0 are always in the model): proc model data=to_forecast_parms outparms=parms_all_Asymm;   dif_&term_rate. =B1*dif_10YrTr_up + B2*dif_10YrTr_down +  B3*lag_dif_10YrTr_up + B4*lag_dif_10YrTr_down +   B5*dif_3MoTr_up + B6*dif_3MoTr_down + B7*lag_dif_3MoTr_up + B8*lag_dif_3MoTr_down + B9*lag_dif_&term_rate. +   Gamma*(A0 + A1*lag_3MoTr + A2*lag_10YrTr + A3*(lag_3MoTr*lag_3MoTr) + A4*(lag_10YrTr*lag_10YrTr) +   A5*(lag_3MoTr*lag_3MoTr*lag_3MoTr) + A6*(lag_10YrTr*lag_10YrTr*lag_10YrTr) - lag_&term_rate.);   fit dif_&term_rate. / dw=1 collin white; run; It's already within a macro where I just hard-coded some of the most significant model forms, "&term_rate." is the macro function parameter which is the variable being modeled in each pass through of the model forms. B1,B2,B3,B4,B5,B6,B7.B8.A1,A2,A3,A4,A5,A6  are the coefficients I need permute each combination of, and the only output I require are the parameter estimates,p-values, R-square/Adj R-square, RMSE, and the value for white's heteroscedasticity test if possible. And yes, I expect it to take a very long time to run, but this is fine for my purposes. As long as it runs in under 24 hours it's fine, even a bit longer won't hurt. I look forward to seeing your suggestion. Thank you! ... View more

Ahhh I see. I was trying different ODS options, I didn't think it would be as simple as just "title;" Thank you very much Cynthia!! ... View more

Gahhhh disregard my last reply. I misunderstood, your suggestion worked. Many thanks Ballard!! ... View more

I tried all 3 of your suggestions, but receive a syntax error for each: keyword sum='';                         _                         22                         200 ERROR 22-322: Syntax error, expecting one of the following: ;, /, ALL, COLPCTN, COLPCTSUM, CSS, CV, KURTOSIS, LCLM, MAX, MEAN,               MEDIAN, MIN, MODE, N, NMISS, P1, P10, P25, P5, P50, P75, P90, P95, P99, PAGEPCTN, PAGEPCTSUM, PCTN, PCTSUM, PROBT,               PRT, Q1, Q3, QRANGE, RANGE, REPPCTN, REPPCTSUM, ROWPCTN, ROWPCTSUM, SKEWNESS, STD, STDDEV, STDERR, SUM, SUMWGT, T,               UCLM, USS, VAR.  ERROR 200-322: The symbol is not recognized and will be ignored. ___________________________________________________________________________________________ mean='';              ____              180 ERROR 180-322: Statement is not valid or it is used out of proper order. ___________________________________________________________________________________________ sum='example text';              ___              180 ERROR 180-322: Statement is not valid or it is used out of proper order. ___________________________________________________________________________________________ Am I mistaken that I should have just added those statements within my proc tabulate statement, before the table statement? ... View more

Thank you yoanbolduc, another helpful suggestion. But many thanks to Astounding!! Once again you have provided me with a very elegant solution that 100% resolves my issue. Can't tell you how much I appreciate it! Thanks, Ryan ... View more

yoanbolduc- Thank you for the suggestion, so you mean such as: data forecasts; set mydata; lag_flat_rate = lag(Flat_Rate); lag_d_flat_rate = lag(d_Flat_Rate); put d_Flat_Rate= lag_flat_rate= lag_d_flat_rate= Flat_Rate= _N_=;   if missing(Flat_rate) then do; d_Flat_Rate=lag_d_flat_rate + /*some arithmetic*/ ; Flat_Rate = d_Flat_Rate + lag_flat_rate; lag_flat_rate = lag(Flat_Rate); lag_d_flat_rate = lag(d_Flat_Rate); end; ?? -That produced the same result as before run; ... View more

Sorry about that, the typo was in my post, not in the code. That isn't the issue ... View more