11.0 RR Command The RR option allows extensive equation specification tests following the procedures suggested by Brown-Durbin-Evans. Tests available include resursive residuals, CUSUM, CUSUMSQ, Quandt- Likelihood, moving regression, and Chow tests. The RR command does not allow lags to be specified on the MODEL statement. If lag variables are desired, they must be built in a B34S DATA step. The ROBUST and REG commands allow for lags to be set on the MODEL sentence. The ROBUST command calculates moving regressions for OL:S, L1 and MINIMAX estimators BUT is substantially slower than the RR command. The ROBUST command uses modified routines originally obtained from version 8 of the IMSL library. Note that the OLSQ command under the MATRIX procedure contains recursive residual capability and well as the ability to produce L1 and MINIMAX results. References for the RR command include: -Brown, R., J. Durbin and J. Evans, "Techniques for Testing the Constancy of Regression Relationships over Time," in Journal of Royal Statistical Society Series B, Vol. 37, 1975. - Dufour, Jean-Marie,"Recursive Stability Analysis of Linear Regression Relationships," Journal of Econometrics, Vol. 19, 1982, pp. 31-76. For OLS models, see REGRESSION, REG, ROBUST and QR commands. The general form of the RR command is: B34SEXEC RR options parameters$ MODEL Yvar = Xvar1 Xvar2 $ BISPEC options parameters$ TRISPEC options parameters$ POLYSPEC options parameters$ B34SEEND$ MODEL is a required sentence. If the NTEST parameters in the RR sentence are omitted, only OLS results are given. RR options: SORTF - Sorts data against each variable listed in NTEST card. SORTB - Reverse sorts data. This option is used to test for the number of observations to drop off the beginning of the dataset. NOINT - A constant will not be estimated. CUSUM10 - Sets 10% bounds on CSUM plot. CUSUM5 - Sets 5% bounds on CSUM plot. CUSUM1 - Sets 1% bounds on CSUM plot. This is the default. Note that the 1% bounds are the widest. ACOV - Same as WHITE command. WHITE - If set uses the White (1980) formula to calculate the SE. For further detail see Greene (1993) page 391. This option is similar to the SAS ACOV option on SAS PROC REG or the the RATS ROBUSTERRORS on the RATS command LINREG. The command ACOV can be used in place of WHITE. RR parameters: NTEST=n1 Instructs the RR command to use the first n1 right hand variables on the MODEL sentence to perform RR equation stability tests. IBEGIN=n2 Sets the observation number for the first time observation. Default=1. IEND=n3 Sets the last time period. Default = last observation in the current dataset. IRB=n4 Sets the maximum order of the recursive residual. Default=1. IRRPT=key If IRRPT is not specified, the recursive residual (RR) will not be plotted. If key=PLOTO, the RR will be plotted aginst order. If key=PLOTX, the RR will be plotted against each X variable. IRRLS=key If IRRLS is not specified, the recursive residuals (RR) will not be listed or saved. If key=LIST, the RR will be listed. If key=LISTPE, the RR will be listed and punched on unit 37 in format 5E16.8. If key=LISTPA, the RR will be listed and saved on unit 44 in SCA FSAVE format in file RRES. Data will also be saved. If key=SCAP, the RR will be saved on unit 44 in SCA FSAVE format in file RRES. Data will also be saved. Note: Recursive residuals of order greater than one will have missing values placed at the end of the series. IBCPT=key If IBCPT is not specified, the recursive B coefficient will not be plotted. If key=PLOTO, the recursive B coefficient will be plotted against order. If key=PLOTX, the recursive B coefficient will be plotted against X. IBCLS=key If IBCLS is not specified, the recursive B coefficient will not be listed, plotted or saved. If key=LIST, the recursive B coefficient will be listed. If key=LISTP, the recursive B coefficient will be listed and saved on unit 44 in SCA FSAVE format in file RCOEF If key=LISTBSB, the recursive B coefficient and standardized B coefficient will be listed. If key=LISTPBSB, the recursive B coefficient and the standardized B coefficient will be listed and saved on unit 44 in SCA FSAVE files RCOEF and RSCOEF respectively If key=LISTSB, the standardized B coefficients will be listed. If key=SCAP, the recursive B coefficient will be saved on unit 44 in SCA FSAVE format in file RCOEF and plotted using line printer plot. If key=SCASBP, the standardized B coefficient will be saved on unit 44 in SCA FSAVE format in file RSCOEF and plotted using line printer plot. If key=SCABBP, both the recursive residual and the standardized residuals are saved in SCA FSAVE format and plotted using line printer plot. INOPUL=n5 Sets number of standardized B coefficient differences calculated. The value n5 must be LE number of series in NTEST (n1) sentence. Default=0. INOPLT=key If INOPLT is not specified, the standardized B coefficients will not be be plotted. If key=PLOTO, the standardized B coefficients will be plotted against order. If key=PLOTX, the standardized B coefficients will be plotted against X. ICUM=key If ICUM is not specified, the CUSUM test is not performed. If key=LIST, the CUSUM test is listed. If key=PLOTO, the CUSUM test is plotted against order. If key=PLOTX, the CUSUM test is plotted against X. If key=PLOTOLS, the CUSUM test is plotted against order, listed and saved in SCA FSAVE file. If key=PLOTXLS, the CUSUM test is plotted against X, listed and saved in SCA FSAVE file. If key=PLOTOL the CUSUM test is plotted against order and listed. If key=PLOTXL the CUSUM test is plotted against X and listed. If key=PLOTOS, the CUSUM test is plotted against order and saved in SCA FSAVE file. If key=PLOTXS, the CUSUM test is plotted against X and saved in SCA FSAVE file. Note: If FSAVE file save performed. SCA savefile will be CUSUMi and variables saved will be OBS, L10, L5, L1, CUSUM, U1, U5, U10 for the CUSUN test for order i. ICUMSQ=key If ICUMSQ is not specified, the CUMSUMSQ test is not performed. If key=LIST. the CUSUMSQ test is listed. If key=PLOTO, the CUSUMSQ test is plotted against order. If key=PLOTX, the CUSUMSQ test is plotted against X. If key=PLOTOLS, the CUSUMSQ test is plotted against order, listed and saved in SCA FSAVE file. If key=PLOTXLS, the CUSUMSQ test is plotted against X, listed and saved in SCA FSAVE file. If key=PLOTOL the CUSUMSQ test is plotted against order and listed. If key=PLOTXL the CUSUMSQ test is plotted against X and listed. If key=PLOTOS, the CUSUMSQ test is plotted against order and saved in SCA FSAVE file. If key=PLOTXS, the CUSUMSQ test is plotted against X and saved in SCA FSAVE file. Note: If FSAVE file save performed. SCA savefile will be CUSUMSQi and variables saved will be OBS, L, CUSUMSQ, BASE, U. For this release of B34S PLOTO** = PLOTA**. IMOVE=n6 Sets the number of periods for autocorrelation of the recursive residual. The maximum allowed = 60. ICHOW=key If ICHOW is not specified, the Chow test is not performed. If key=CAL, the Chow test is performed. If key=CALPRINT, the Chow test is performed and the OLS subsample results are printed. If key=PRINT, the OLS subsample results are printed but the Chow test is not performed. IOBSCH=n7 Sets the number of observations per subsample for the Chow test. IOBSCH must be greater than the number of right hand variables. The number of Chow subsamples calculated = (IEND-IBEGIN+1)/IOBSCH. If IOBSCH will not go in evenly, the remainder is added to the last subsample. IWN=n8 Sets the maximum order of the Wilcox test on the IRB recursive residual sets. The default = 1. The Wilcox test will be performed on Z(i)=W(i) * W(I+k) for k going from 1 to n8. IADD=n9 Sets the number of additional observations to add to the base. The default is 1. The IADD parameter is used if the researcher knows that, due to a dummy variable, a full rank matrix cannot be obtrained for the first n8-1 observations in the dataset. If the IADD parameter is not specified, the B34S RR procedure will have to test each prospective base for rank. IQUANT=key If IQUANT is not specified, Quandt's Log-Likelihood Ratio Test (QLRT) will not be performed. If key=PLOT, the QLRT test will be plotted. If key=LIST, the QLRT will be listed. If key=PLIST, the QLRT will be listed and plotted. (LISTP is a synonym for PLIST). If key=PPLIST, the QLRT will be listed, plotted and saved on unit 44 in SCA FSAVE format in file RRQUANT. Series names are OBSNUM, QUANDTLR, SIGMA1, SIGMA2, TSUMSQ. Note that TSUMSQ is a constant. If key=SCAP, the QLRT will be plotted and saved on unit 44 in SCA FSAVE format in file RRQUANT. Series names are OBSNUM, QUANDTLR, SIGMA1, SIGMA2, TSUMSQ. Note that TSUMSQ is a constant. IMORR=key If IMORR is not specified, moving regressions are not calculated. If key=MTEST, m1 (first step ahead forecast error), m2 (first step behind forecast error), m (m=m1+m2) and m3 (constant interval form of m1) are calculated. If key=LIST, moving coefficients are listed as well the test statistics for MTEST. Values listed are not normalized. Beginning values set to zero. Listed M values show how the sum is proceeding. If key=PUNCH, moving coefficients and MTEST statistics are placed on unit 44 in SCA FSAVE format. File name = MCOEF. Data in form: #obs in reg (NOOBSREG) ., Obs beginning (NOBBEGIN), coef 1- K, M1, M2, M, M3, RSUMSQ. (See notes for LIST). Another SCA FSAVE file MVALUES containing NOOBSREG NOBBEGIN M1 M2 M M3 is also saved. These represent normalized M values. If key=LPUNCH, option LIST and PUNCH are performed. If key=SOUTPUT, the test statistics are listed only after each set of moving regressions. This option saves paper (i.e. suppressed output). Note that the SCA FSAVE file produced by PUNCH or LPUNCH may require further processing to extract sets of moving coefficients for different NOOBSREG values. NBEGIN=n10 Sets the number of observations to use to begin moving regressions. The default is the number of variables on the right hand side of the MODEL sentence. MMRNOB=n11 Sets the number of observations in the last set of moving regressions. n11 must be GT than n10 and LT number of observations in the dataset. If MMRNOB is set outside this range, if defaults to NBEGIN. INCR=n12 Sets the increment for moving regressions. The default = 1. If NBEGIN=7 and MMRNOB=14 and INCR=2, sets of moving regressions will be attempted for # of observations 7, 9, 11, 13. MODEL sentence. MODEL Y = X1 X2 X3 X4 $ The MODEL sentence lists the left hand variable and the right hand variables. A constant will be added to the variable list unless the NOINT option is given. The NTEST parameter in the RR sentence sets how many variables to the right of the = sign on the MODEL card are to be used for stability tests of the coefficients of the the variables listed on the right hand side of the MODEL sentence. The maximum number of variables that can be listed on the right hand side of the MODEL sentence is 69 (including the constant). BISPEC sentence. The BISPEC sentence performs various nonlinearity, gaussianity and matringale tests suggested by Hinich. The form of the BISP sentence in the BTIDEN, BTEST and MARS commands is the same. To save space, detail for this sentence is only given under the BTIDEN command help file. If the BISPEC sentence is given with no options or parameters, gaussianity and nonlinearity tests will be performed using default settings. The setting BISPEC IAUTO ITURNO $ will perform tests for gaussianity and nonlinearity over a grid of admissable values for the bandwidth. TRISPEC sentence The TRISPEC command performs 4th order nonlinearity tests suggested by Hinich. Further detail on this sentence is listed under the BTIDEN command. POLYSPEC sentence The POLYSPEC command performs various nonlinearity tests suggested by Hinich within the sample. Further detail on this sentence is listed under the BTIDEN command. Sample RR setup for recursive residual analysis. Where the data is first sorted against X1, and next against X2. If the SORTF option had not been given, the data would have not been sorted. B34SEXEC RR ICUM=PLOTO SORTF IRB=2 IRRPT=PLOTO IRRLS=LIST IBCPT=PLOTO IBCLS=LIST ICHOW=CALPRINT IOBSCH=25 IQUANT=PLOT NTEST=2 $ MODEL Y = X1 X2 X3 $ B34SEEND $ Sample RR setup for a simple regression with nonlinearity tests. Recursive residual test are not given because the TEST parameter has not been specified. B34SEXEC RR $ MODEL Y = X1 X2 X3 $ BISPEC IAUTO ITURNO$ B34SEEND$