TSP regression output, 1978-2007

In 1978 when Bronwyn H. Hall and Robert E. Hall prepared the rewritten version of the TSP 3.3 User's Guide on the new mainframe laser printer, they included TSP output from an illustrative model, which included OLSQ, INST, LSQ, ANALYZ, FIML and SIML. The OLSQ output has been updated in subsequent editions of the TSP User's Guide, which allows us to compare the regression output as it has been modified slightly over the years.

1978: TSP 3.3

Diagnostics are left adjusted, not aligned on = sign. Output for TSP 2.8 is about the same. EQUATION 1 ************ ORDINARY LEAST SQUARES DEPENDENT VARIABLE: CONS SUM OF SQUARED RESIDUALS = 3958.15 STANDARD ERROR OF THE REGRESSION = 12.5828 MEAN OF DEPENDENT VARIABLE = 519.032 STANDARD DEVIATION = 147.459 R-SQUARED = 0.9930 ADJUSTED R-SQUARED = 0.9927 F-STATISTIC( 1., 25.) = 3545.80 LOG OF LIKELIHOOD FUNCTION = -105.645 NUMBER OF OBSERVATIONS = 27. SUM OF RESIDUALS = .366211E-02 DURBIN-WATSON STATISTIC (ADJ. FOR 0. GAPS) = 0.4617 RIGHT-HAND ESTIMATED STANDARD T- VARIABLE COEFFICIENT ERROR STATISTIC C -17.8024 9.33493 -1.90707 GNP .633919 .106458E-01 59.5467 ESTIMATE OF VARIANCE-COVARIANCE MATRIX OF ESTIMATED COEFFICIENTS C GNP .............................. C . 87.1409 -.959755E-01 GNP . -.959755E-01 .113332E-03 1 2

1983: TSP 4.0

Diagnostics are right adjusted, aligned on the = sign. Sum of residuals is deleted. EQUATION 1 ************ METHOD OF ESTIMATION = ORDINARY LEAST SQUARES DEPENDENT VARIABLE: CONS SUM OF SQUARED RESIDUALS = 3958.16 STANDARD ERROR OF THE REGRESSION = 12.5828 MEAN OF DEPENDENT VARIABLE = 519.033 STANDARD DEVIATION = 147.459 R-SQUARED = .992999 ADJUSTED R-SQUARED = .992719 DURBIN-WATSON STATISTIC (ADJ. FOR 0 GAPS) = 0.4617 NUMBER OF OBSERVATIONS = 27 F-STATISTIC( 1, 25) = 3545.79 LOG OF LIKELIHOOD FUNCTION = -105.645 ESTIMATED STANDARD VARIABLE COEFFICIENT ERROR T-STATISTIC C -17.80238 9.334954 -1.907067 GNP .6339192 .1064578D-01 59.54652 VARIANCE COVARIANCE OF ESTIMATED COEFFICIENTS C GNP -------------------------- C 87.14136 GNP -0.09598 0.00011

1987: TSP 4.1

Drop the VCOV matrix and remove the "GAPS" label from the Durbin-Watson. EQUATION 1 ************ METHOD OF ESTIMATION = ORDINARY LEAST SQUARES DEPENDENT VARIABLE: CONS SUM OF SQUARED RESIDUALS = 3958.17 STANDARD ERROR OF THE REGRESSION = 12.5828 MEAN OF DEPENDENT VARIABLE = 519.033 STANDARD DEVIATION = 147.459 R-SQUARED = 0.992999 ADJUSTED R-SQUARED = 0.992719 DURBIN-WATSON STATISTIC = 0.4617 F-STATISTIC( 1, 25) = 3545.79 LOG OF LIKELIHOOD FUNCTION = -105.645 NUMBER OF OBSERVATIONS = 27 ESTIMATED STANDARD VARIABLE COEFFICIENT ERROR T-STATISTIC C -17.802 9.3350 -1.9071 GNP 0.63392 0.10646E-01 59.547

1993: TSP 4.2

New 2-column format for diagnostics, with mixed upper and lowercase text. Add new statistic: Schwarz B.I.C. (normalized) Add Current sample to header and move Number of obs to header. Equation 1 ************ Method of estimation = Ordinary Least Squares Dependent variable: CONS Current sample: 1949 to 1975 Number of observations: 27 Mean of dependent variable = 519.033 Adjusted R-squared = .992719 Std. dev. of dependent var. = 147.459 Durbin-Watson statistic = .461739 Sum of squared residuals = 3958.17 F-statistic (zero slopes) = 3545.79 Variance of residuals = 158.327 Schwarz Bayes. Info. Crit. = 5.23184 Std. error of regression = 12.5828 log of likelihood function = -105.645 R-squared = .992999 Estimated Standard Variable Coefficient Error t-statistic C -17.8024 9.33496 -1.90707 GNP .633919 .010646 59.5465

1997: TSP 4.4 and later

Add new default tests: LM het test, Jarque-Bera, RESET2, switch to unnormalized SBIC, add p-values for all tests, including Durbin-Watson. Reduce size of labels to make room for p-values.
Note: shown are lower and upper bounds on p-value for Durbin-Watson, which in this case imply that the p-value is less than .0005 . Equation 1 ============ Method of estimation = Ordinary Least Squares Dependent variable: CONS Current sample: 1949 to 1975 Number of observations: 27 Mean of dep. var. = 519.033 LM het. test = .783303 [.376] Std. dev. of dep. var. = 147.459 Durbin-Watson = .461739 [.000,.000] Sum of squared residuals = 3958.17 Jarque-Bera test = 3.29859 [.192] Variance of residuals = 158.327 Ramsey's RESET2 = 16.7953 [.000] Std. error of regression = 12.5828 F (zero slopes) = 3545.79 [.000] R-squared = .992999 Schwarz B.I.C. = 108.941 Adjusted R-squared = .992719 Log likelihood = -105.645 Estimated Standard Variable Coefficient Error t-statistic P-value C -17.8024 9.33496 -1.90707 [.068] GNP .633919 .010646 59.5465 [.000]