Welcome to SHAZAM - Version 9.0 - APR 2003 SYSTEM=WIN-98 PAR= 2000 CURRENT WORKING DIRECTORY IS: C:\ECONOM~1\BERNDT\CHAP8.DAT |_* CHAPTER 8 - CAUSALITY AND SIMULTANEITY NETWEEN ADVERTISING AND SALES |_* |_* |_* Exercise 7 - Evaluating the Effects of the Cigarette Broadcast Ban |_* |_* Set the beginning of the sample period (1910) and set for annual data, and read in the data. |_TIME 1930 1 |_SAMPLE 1930.0 1978.0 |_READ (CIGAD) YEAR SALES SALESPC INCPC RPRICE CIGTOB TOBPC PRTOB DF ASTOCK / SKIPLINES=1 UNIT 88 IS NOW ASSIGNED TO: CIGAD 10 VARIABLES AND 49 OBSERVATIONS STARTING AT OBS 1 Read in the second block of data using the NOREWIND option. |_READ (CIGAD) TIME ADV TOBADV REALAD TPERCIG ASTOCK1 ASTOCK2 F L LNI / SKIPLINES=1 NOREWIND 10 VARIABLES AND 49 OBSERVATIONS STARTING AT OBS 1 |_* Part (a) Prepare the data for estimation of equation (8.60). |_GENR LSALESPC=LOG(SALESPC) |_GENR LINCPC=LOG(INCPC) |_GENR LRPRICE=LOG(RPRICE) |_GENR LASTOCK=LOG(ASTOCK) |_GENR D53=(YEAR.GE.1953) |_GENR D64=(YEAR.GE.1964) |_SAMPLE 1930.0 1970.0 The FC command generates ForeCasts based on an immediately preceding OLS command. However, the variable used to store the forecasts (here LNCHAT) must be defined previously by a DIM command. See SHAZAM Manual, p. 206 and 405. |_DIM LNCHAT 49 |_OLS LSALESPC LINCPC LRPRICE LASTOCK DF D53 D64 / RSTAT REQUIRED MEMORY IS PAR= 15 CURRENT PAR= 2000 OLS ESTIMATION 41 OBSERVATIONS DEPENDENT VARIABLE= LSALESPC ...NOTE..SAMPLE RANGE SET TO: 1, 41 R-SQUARE = 0.9599 R-SQUARE ADJUSTED = 0.9529 VARIANCE OF THE ESTIMATE-SIGMA**2 = 0.78883E-02 STANDARD ERROR OF THE ESTIMATE-SIGMA = 0.88816E-01 SUM OF SQUARED ERRORS-SSE= 0.26820 MEAN OF DEPENDENT VARIABLE = 7.8918 LOG OF THE LIKELIHOOD FUNCTION = 44.9301 VARIABLE ESTIMATED STANDARD T-RATIO PARTIAL STANDARDIZED ELASTICITY NAME COEFFICIENT ERROR 34 DF P-VALUE CORR. COEFFICIENT AT MEANS LINCPC 1.2888 0.7274E-01 17.72 0.000 0.950 1.0158 1.1419 LRPRICE -0.72355 0.3044 -2.377 0.023-0.377 -0.1546 -0.4112 LASTOCK 0.32377E-01 0.9797E-01 0.3305 0.743 0.057 0.0392 0.0028 DF -0.11411 0.6207E-01 -1.838 0.075-0.301 -0.0840 -0.0012 D53 0.38024E-01 0.7636E-01 0.4979 0.622 0.085 0.0467 0.0021 D64 -0.11489 0.6727E-01 -1.708 0.097-0.281 -0.1070 -0.0025 CONSTANT 2.1159 1.505 1.406 0.169 0.234 0.0000 0.2681 DURBIN-WATSON = 0.6658 VON NEUMANN RATIO = 0.6825 RHO = 0.54254 RESIDUAL SUM = -0.60764E-13 RESIDUAL VARIANCE = 0.78883E-02 SUM OF ABSOLUTE ERRORS= 2.5617 R-SQUARE BETWEEN OBSERVED AND PREDICTED = 0.9599 RUNS TEST: 11 RUNS, 23 POS, 0 ZERO, 18 NEG NORMAL STATISTIC = -3.2749 The 1953 reports do not seem to have had any effect on cigarette consumption. The U.S. Surgeon-General's 1964 report and the Fairness Doctrine each appear to have lowered per capita cigarette consumption by about 11 percent, but in neither case can we reject the null hypothesis that there was no effect on cigarette consumption at the 5% level of significance. The logarithms of income and price are statistically significant, but the log of advertising goodwill is not. |_FC /BEG=1971.0 END=1978.0 LIST PREDICT=LNCHAT REQUIRED MEMORY IS PAR= 13 CURRENT PAR= 2000 DEPENDENT VARIABLE = LSALESPC 8 OBSERVATIONS REGRESSION COEFFICIENTS 1.28883679838 -0.723550773148 0.323765211981E-01 -0.114111687667 0.380238103413E-01 -0.114886348620 2.11591338597 OBS. OBSERVED PREDICTED CALCULATED STD. ERROR NO. VALUE VALUE RESIDUAL 42 8.1912 8.3268 -0.13563 0.104 * I 43 8.1928 8.3856 -0.19278 0.108 * I 44 8.2191 8.4660 -0.24694 0.107 * I 45 8.2185 8.4632 -0.24473 0.103 * I 46 8.2185 8.4438 -0.22524 0.101 * I 47 8.2126 8.5227 -0.31017 0.101 * I 48 8.2033 8.6868 -0.48347 0.104 * I 49 8.1865 8.7624 -0.57597 0.110 * I SUM OF ABSOLUTE ERRORS= 2.4149 R-SQUARE BETWEEN OBSERVED AND PREDICTED = 0.0446 MEAN ERROR = -0.30187 SUM-SQUARED ERRORS = 0.88885 MEAN SQUARE ERROR = 0.11111 MEAN ABSOLUTE ERROR= 0.30187 ROOT MEAN SQUARE ERROR = 0.33333 MEAN SQUARED PERCENTAGE ERROR= 16.523 THEIL INEQUALITY COEFFICIENT U =28.197 DECOMPOSITION PROPORTION DUE TO BIAS = 0.82014 PROPORTION DUE TO VARIANCE = 0.14141 PROPORTION DUE TO COVARIANCE = 0.38451E-01 DECOMPOSITION PROPORTION DUE TO BIAS = 0.82014 PROPORTION DUE TO REGRESSION = 0.17846 PROPORTION DUE TO DISTURBANCE = 0.14022E-02 Exponentiate the forecasts LNCHAT, and compare with actual sales SALESPC. |_SAMPLE 1971.0 1978.0 |_GENR FORECAST=EXP(LNCHAT) |_GENR FERROR=FORECAST/SALESPC |_PRINT YEAR SALESPC FORECAST FERROR YEAR SALESPC FORECAST FERROR 1971.000 3609.000 4133.216 1.145252 1972.000 3615.000 4383.620 1.212620 1973.000 3711.000 4750.478 1.280107 1974.000 3709.000 4737.418 1.277276 1975.000 3709.000 4645.976 1.252622 1976.000 3687.000 5027.801 1.363656 1977.000 3653.000 5924.036 1.621691 1978.000 3592.000 6389.635 1.778852 These forecasts suggest that the model is not reliable. |_* Part (b) |_SAMPLE 1930.0 1978.0 |_GENR D71=(YEAR.GE.1971) |_OLS LSALESPC LINCPC LRPRICE LASTOCK DF D53 D64 D71 REQUIRED MEMORY IS PAR= 17 CURRENT PAR= 2000 OLS ESTIMATION 49 OBSERVATIONS DEPENDENT VARIABLE= LSALESPC ...NOTE..SAMPLE RANGE SET TO: 1, 49 R-SQUARE = 0.9497 R-SQUARE ADJUSTED = 0.9411 VARIANCE OF THE ESTIMATE-SIGMA**2 = 0.90286E-02 STANDARD ERROR OF THE ESTIMATE-SIGMA = 0.95019E-01 SUM OF SQUARED ERRORS-SSE= 0.37017 MEAN OF DEPENDENT VARIABLE = 7.9430 LOG OF THE LIKELIHOOD FUNCTION = 50.1694 VARIABLE ESTIMATED STANDARD T-RATIO PARTIAL STANDARDIZED ELASTICITY NAME COEFFICIENT ERROR 41 DF P-VALUE CORR. COEFFICIENT AT MEANS LINCPC 1.2523 0.7678E-01 16.31 0.000 0.931 1.1686 1.1172 LRPRICE -0.18991 0.2656 -0.7150 0.479-0.111 -0.0451 -0.1077 LASTOCK -0.15470 0.8394E-01 -1.843 0.073-0.277 -0.2096 -0.0155 DF -0.12885 0.6611E-01 -1.949 0.058-0.291 -0.0911 -0.0012 D53 0.16471 0.7054E-01 2.335 0.025 0.343 0.2122 0.0110 D64 -0.74220E-01 0.7060E-01 -1.051 0.299-0.162 -0.0883 -0.0029 D71 -0.31164 0.6176E-01 -5.046 0.000-0.619 -0.2974 -0.0064 CONSTANT 0.43359E-01 1.408 0.3079E-01 0.976 0.005 0.0000 0.0055 Price is no longer statistically significant; cigarette advertising is seen to LOWER cigarette consumption, and the 1953 reports are associated with an INCREASE (statistically significant) in cigarette consumption. This agrees with the conclusions of Schneider et al. |_* Part (c) If the logic of their model is correct, the coefficient on the log-fitted value of the regression should be unity by definition. |_GENR LCIGTOB=LOG(CIGTOB) |_NL 1 / NCOEF=2 ...NOTE..SAMPLE RANGE SET TO: 1, 49 |_EQ LCIGTOB = A + LOG(1 + 2*(YC/INCPC)) - 2*YC/INCPC |_END 2 VARIABLES IN 1 EQUATIONS WITH 2 COEFFICIENTS 49 OBSERVATIONS REQUIRED MEMORY IS PAR= 32 CURRENT PAR= 2000 COEFFICIENT STARTING VALUES A 1.0000 YC 1.0000 100 MAXIMUM ITERATIONS, CONVERGENCE = 0.100000E-04 INITIAL STATISTICS : TIME = 0.0500 SEC. ITER. NO. 0 FUNCT. EVALUATIONS 1 LOG-LIKELIHOOD FUNCTION= -127.4977 COEFFICIENTS 1.000000 1.000000 GRADIENT 14.97624 -0.5255033E-04 INTERMEDIATE STATISTICS : TIME = 0.0500 SEC. ITER. NO. 15 FUNCT. EVALUATIONS 32 LOG-LIKELIHOOD FUNCTION= 52.59628 COEFFICIENTS 4.632712 634.0620 GRADIENT 155.4056 -0.3478359 FINAL STATISTICS : TIME = 0.0500 SEC. ITER. NO. 25 FUNCT. EVALUATIONS 42 LOG-LIKELIHOOD FUNCTION= 64.84511 COEFFICIENTS 4.575300 551.0631 GRADIENT 0.1778089E-05 0.1333265E-08 MAXIMUM LIKELIHOOD ESTIMATE OF SIGMA-SQUARED = 0.41501E-02 SUM OF SQUARED ERRORS = 0.20336 GTRANSPOSE*INVERSE(H)*G STATISTIC - = 0.26837E-14 COEFFICIENT ST. ERROR T-RATIO A 4.5753 0.17810E-01 256.89 YC 551.06 16.532 33.333 |_END The constant term differs from that reported by Schneider et al., which is -0.0299. This appears to be because our data represent the variable CIGTOB as a percentage, whereas Schneider seem to have used raw proportions. The difference in the logarithms of the two series is log(100)=4.6052. |_* Part (d) |_STAT LNI LINCPC /PCOR NAME N MEAN ST. DEV VARIANCE MINIMUM MAXIMUM LNI 49 -0.34043 0.22736 0.51694E-01 -1.0211 -0.99230E-01 LINCPC 49 7.0858 0.36521 0.13338 6.2689 7.7125 CORRELATION MATRIX OF VARIABLES - 49 OBSERVATIONS LNI 1.0000 LINCPC 0.95954 1.0000 LNI LINCPC Given the high collinearity between LNI and LLINCPC, it may be difficult to obtain precise estimates of the coefficients when both variables are included as regressors. |_GENR LINCPC2=LINCPC**2 |_STAT LNI LINCPC LINCPC2 /PCOR NAME N MEAN ST. DEV VARIANCE MINIMUM MAXIMUM LNI 49 -0.34043 0.22736 0.51694E-01 -1.0211 -0.99230E-01 LINCPC 49 7.0858 0.36521 0.13338 6.2689 7.7125 LINCPC2 49 50.339 5.1109 26.122 39.299 59.483 CORRELATION MATRIX OF VARIABLES - 49 OBSERVATIONS LNI 1.0000 LINCPC 0.95954 1.0000 LINCPC2 0.95090 0.99956 1.0000 LNI LINCPC LINCPC2 |_OLS LSALESPC LINCPC LINCPC2 LRPRICE LASTOCK DF D53 D64 D71 /COEF=BETA REQUIRED MEMORY IS PAR= 19 CURRENT PAR= 2000 OLS ESTIMATION 49 OBSERVATIONS DEPENDENT VARIABLE= LSALESPC ...NOTE..SAMPLE RANGE SET TO: 1, 49 R-SQUARE = 0.9513 R-SQUARE ADJUSTED = 0.9415 VARIANCE OF THE ESTIMATE-SIGMA**2 = 0.89601E-02 STANDARD ERROR OF THE ESTIMATE-SIGMA = 0.94658E-01 SUM OF SQUARED ERRORS-SSE= 0.35840 MEAN OF DEPENDENT VARIABLE = 7.9430 LOG OF THE LIKELIHOOD FUNCTION = 50.9609 VARIABLE ESTIMATED STANDARD T-RATIO PARTIAL STANDARDIZED ELASTICITY NAME COEFFICIENT ERROR 40 DF P-VALUE CORR. COEFFICIENT AT MEANS LINCPC 4.6455 2.962 1.568 0.125 0.241 4.3349 4.1442 LINCPC2 -0.25112 0.2191 -1.146 0.259-0.178 -3.2793 -1.5915 LRPRICE -0.32205 0.2886 -1.116 0.271-0.174 -0.0765 -0.1826 LASTOCK -0.10339 0.9485E-01 -1.090 0.282-0.170 -0.1401 -0.0104 DF -0.99210E-01 0.7075E-01 -1.402 0.169-0.216 -0.0702 -0.0009 D53 0.15639 0.7065E-01 2.214 0.033 0.330 0.2015 0.0104 D64 -0.37719E-01 0.7720E-01 -0.4886 0.628-0.077 -0.0449 -0.0015 D71 -0.23068 0.9369E-01 -2.462 0.018-0.363 -0.2201 -0.0047 CONSTANT -10.827 9.588 -1.129 0.266-0.176 0.0000 -1.3631 The regression coefficients are saved in the vector BETA. The coefficient on LINCPC is BETA(1) and that on LINCPC2 is BETA(2). Income elasticity is dlog(Cons)/dlog(Income)=BETA(1)+2*BETA(2)*log(Income). |_GENR INCELAS=BETA(1)+2*BETA(2)*LINCPC |_STAT INCELAS NAME N MEAN ST. DEV VARIANCE MINIMUM MAXIMUM INCELAS 49 1.0867 0.18342 0.33644E-01 0.77197 1.4970 |_PRINT YEAR INCELAS YEAR INCELAS 1930.000 1.354508 1931.000 1.390435 1932.000 1.479624 1933.000 1.497015 1934.000 1.441669 1935.000 1.380228 1936.000 1.332161 1937.000 1.289426 1938.000 1.327659 1939.000 1.291697 1940.000 1.252085 1941.000 1.178779 1942.000 1.124381 1943.000 1.092140 1944.000 1.086941 1945.000 1.110265 1946.000 1.125061 1947.000 1.141617 1948.000 1.130856 1949.000 1.143073 1950.000 1.105530 1951.000 1.075180 1952.000 1.070763 1953.000 1.058218 1954.000 1.077444 1955.000 1.051120 1956.000 1.051943 1957.000 1.055761 1958.000 1.071293 1959.000 1.047764 1960.000 1.043774 1961.000 1.042733 1962.000 1.019703 1963.000 1.006741 1964.000 0.9842069 1965.000 0.9622139 1966.000 0.9351094 1967.000 0.9285941 1968.000 0.9103683 1969.000 0.9027053 1970.000 0.9102796 1971.000 0.8990185 1972.000 0.8689858 1973.000 0.8461371 1974.000 0.8628658 1975.000 0.8762581 1976.000 0.8510817 1977.000 0.7931637 1978.000 0.7719702 While the estimated income elasticity becomes more reasonable toward the end of the sample, this is not a particularly good specification, and is not a viable alternative to that in (b). The price term is not significant, the coefficient on LASTOCK has the wrong sign (although it is not signaficant), and the coefficients on DF and D64 are not significant. The high (0.99956) correlation between LINCPC and LINCPC2 also suggests that the coefficents for these variables (and therefore the elasticity estimates) are likely very imprecisely measured. |_* Part (e) |_NL 1 /NCOEF=10 ...NOTE..SAMPLE RANGE SET TO: 1, 49 |_EQ LSALESPC=BY*LINCPC+CON+BPR*LRPRICE+BI*LNI+BA*LOG(ASTOCK1+r*ASTOCK2)+BDF*DF+BF*F+BL*L+BTPC*LOG(TPERCIG) |_COEF BY 0.462 CON 2.243 BPR -1.218 BI 0.971 BA 0.046 R 0.264 BDF -0.075 BF -0.0021 BL -0.0235 BTPC -1.386 10 VARIABLES IN 1 EQUATIONS WITH 10 COEFFICIENTS 49 OBSERVATIONS REQUIRED MEMORY IS PAR= 41 CURRENT PAR= 2000 COEFFICIENT STARTING VALUES BY 0.46200 CON 2.2430 BPR -1.2180 BI 0.97100 BA 0.46000E-01 R 0.26400 BDF -0.75000E-01 BF -0.21000E-02 BL -0.23500E-01 BTPC -1.3860 100 MAXIMUM ITERATIONS, CONVERGENCE = 0.100000E-04 INITIAL STATISTICS : TIME = 0.0000 SEC. ITER. NO. 0 FUNCT. EVALUATIONS 1 LOG-LIKELIHOOD FUNCTION= 51.66360 COEFFICIENTS 0.4620000 2.243000 -1.218000 0.9710000 0.4600000E-01 0.2640000 -0.7500000E-01 -0.2100000E-02 -0.2350000E-01 -1.386000 GRADIENT -733.8386 -97.55878 -442.8933 2.986384 -89.70877 -3.504481 -16.38617 -6746.643 -599.3144 603.1006 INTERMEDIATE STATISTICS : TIME = 0.0000 SEC. ITER. NO. 15 FUNCT. EVALUATIONS 22 LOG-LIKELIHOOD FUNCTION= 55.68616 COEFFICIENTS 1.102985 -0.4720713 -1.195243 0.1361023 0.6584351E-01 0.4193096 -0.9398522E-01 -0.2719374E-02 -0.2691339E-01 -1.018878 GRADIENT 5.970906 0.8311281 3.779913 -0.3212313 0.8759897 0.1019667E-01 0.2390204 59.54645 6.974769 -5.154633 FINAL STATISTICS : TIME = 0.0000 SEC. ITER. NO. 21 FUNCT. EVALUATIONS 28 LOG-LIKELIHOOD FUNCTION= 55.68665 COEFFICIENTS 1.102630 -0.4667122 -1.196664 0.1352938 0.6532288E-01 0.4025064 -0.9393025E-01 -0.2702838E-02 -0.2682475E-01 -1.019414 GRADIENT 0.2049775E-03 0.2935557E-04 0.1353581E-03 -0.1115729E-04 0.6368459E-04 -0.2539258E-05 0.7539236E-05 0.1776418E-02 -0.3677532E-03 -0.1777947E-03 MAXIMUM LIKELIHOOD ESTIMATE OF SIGMA-SQUARED = 0.60312E-02 SUM OF SQUARED ERRORS = 0.29553 GTRANSPOSE*INVERSE(H)*G STATISTIC - = 0.26805E-11 COEFFICIENT ST. ERROR T-RATIO BY 1.1026 0.33338 3.3074 CON -0.46671 3.9195 -0.11907 BPR -1.1967 0.20447 -5.8525 BI 0.13529 0.40120 0.33722 BA 0.65323E-01 0.66850E-01 0.97716 R 0.40251 0.72744 0.55332 BDF -0.93930E-01 0.50136E-01 -1.8735 BF -0.27028E-02 0.26665E-02 -1.0136 BL -0.26825E-01 0.64015E-02 -4.1904 BTPC -1.0194 0.65913 -1.5466 |_END |_TEST BY=.462 TEST VALUE = 0.64063 STD. ERROR OF TEST VALUE 0.33338 ASYMPTOTIC NORMAL STATISTIC = 1.9216271 P-VALUE= 0.05465 WALD CHI-SQUARE STATISTIC = 3.6926509 WITH 1 D.F. P-VALUE= 0.05465 UPPER BOUND ON P-VALUE BY CHEBYCHEV INEQUALITY = 0.27081 The estimate of income elasticity is still quite large (1.10), although the null hypothesis that it is 0.462 cannot be rejected (but just barely). The 0.135 estimate on the coefficient of LNI is quite different from unity (although unity cannot be rejected). The estimate of 0.40 on r is very imprecise. Imposing an income elasticity of 0.462: |_NL 1 /NCOEF=9 ...NOTE..SAMPLE RANGE SET TO: 1, 49 |_EQ LSALESPC=0.462*LINCPC+CON+BPR*LRPRICE+BI*LNI+BA*LOG(ASTOCK1+r*ASTOCK2)+BDF*DF+BF*F+BL*L+BTPC*LOG(TPERCIG) |_COEF CON 2.243 BPR -1.218 BI 0.971 BA 0.046 R 0.264 BDF -0.075 BF -0.0021 BL -0.0235 BTPC -1.386 10 VARIABLES IN 1 EQUATIONS WITH 9 COEFFICIENTS 49 OBSERVATIONS REQUIRED MEMORY IS PAR= 40 CURRENT PAR= 2000 COEFFICIENT STARTING VALUES CON 2.2430 BPR -1.2180 BI 0.97100 BA 0.46000E-01 R 0.26400 BDF -0.75000E-01 BF -0.21000E-02 BL -0.23500E-01 BTPC -1.3860 100 MAXIMUM ITERATIONS, CONVERGENCE = 0.100000E-04 INITIAL STATISTICS : TIME = 0.0000 SEC. ITER. NO. 0 FUNCT. EVALUATIONS 1 LOG-LIKELIHOOD FUNCTION= 51.66360 COEFFICIENTS 2.243000 -1.218000 0.9710000 0.4600000E-01 0.2640000 -0.7500000E-01 -0.2100000E-02 -0.2350000E-01 -1.386000 GRADIENT -97.55878 -442.8933 2.986384 -89.70877 -3.504481 -16.38617 -6746.643 -599.3144 603.1006 INTERMEDIATE STATISTICS : TIME = 0.0600 SEC. ITER. NO. 15 FUNCT. EVALUATIONS 24 LOG-LIKELIHOOD FUNCTION= 53.77321 COEFFICIENTS 2.703843 -1.202153 0.8928306 0.4268248E-01 0.2582170 -0.7615237E-01 -0.1801626E-02 -0.2317936E-01 -1.289629 GRADIENT -0.5886421E-01 -0.2704514 0.2101233E-04 -0.1606613E-01 0.2537208E-02 -0.2862210E-02 -4.754379 -0.7366880 0.3689903 FINAL STATISTICS : TIME = 0.0600 SEC. ITER. NO. 19 FUNCT. EVALUATIONS 28 LOG-LIKELIHOOD FUNCTION= 53.77322 COEFFICIENTS 2.703585 -1.202314 0.8928164 0.4283092E-01 0.2620470 -0.7611304E-01 -0.1804002E-02 -0.2319625E-01 -1.289785 GRADIENT 0.1409560E-04 0.7075533E-04 0.3896052E-05 -0.6931329E-04 -0.5011837E-07 0.8626820E-05 0.1031214E-02 0.7154919E-04 -0.9593189E-04 MAXIMUM LIKELIHOOD ESTIMATE OF SIGMA-SQUARED = 0.65212E-02 SUM OF SQUARED ERRORS = 0.31954 GTRANSPOSE*INVERSE(H)*G STATISTIC - = 0.41138E-10 COEFFICIENT ST. ERROR T-RATIO CON 2.7036 3.6546 0.73977 BPR -1.2023 0.20364 -5.9042 BI 0.89282 0.79791E-01 11.189 BA 0.42831E-01 0.56206E-01 0.76203 R 0.26205 0.71576 0.36611 BDF -0.76113E-01 0.49909E-01 -1.5250 BF -0.18040E-02 0.25584E-02-0.70513 BL -0.23196E-01 0.63518E-02 -3.6519 BTPC -1.2898 0.64919 -1.9868 |_END The LNI coefficient BI has increased so that it is now almost unity (with a small standard error relative to the earlier run). While the health information variables are all negative, as expected, only the low-tar variable is significant. ..INPUT FILE COMPLETED..TYPE A NEW COMMAND OR TYPE: STOP Updated November 10, 2008