* PS11.1, using DATA10-5, for fitting time trends in California wage rate PAR 500 TIME 1960 1 SAMPLE 1960. 1994. READ(data10-5) YEAR CAL US * * GENERATE TRANSFORMED VARIABLES * GENR T=TIME(0) PRINT T GENR TSQ=T*T GENR T3=TSQ*T GENR INVT=1/T GENR LTIME=LOG(T) GENR LCAL=LOG(CAL) DIM YHATA 50 YHATB 50 YHATC 50 YHATD 50 YHATE 50 YHATF 50 YHATG 50 * SAMPLE 1960. 1989. OLS CAL T * * Model A - Linear Model. * AUTO CAL T / MAX DROP FC / MAX PREDICT=YHATA IBLUP BEG=1989. END=1994. * * Regress actual against predicted. * SAMPLE 1990. 1994. OLS CAL YHATA * * Compute prediction error, error sum of squares, and selection criteria. * GENR UHATA=CAL-YHATA GENR MA=(100*ABS(UHATA)/CAL) STAT MA / MEAN=MAPEA GENR UHATA2=UHATA*UHATA GEN1 ESSA=SUM(UHATA2) * SAMPLE 1960. 1989. * * Model B - Quadratic Model. * AUTO CAL T TSQ / MAX DROP FC / MAX PREDICT=YHATB IBLUP BEG=1989. END=1994. * SAMPLE 1990. 1994. OLS CAL YHATB * * Compute prediction error, error sum of squares, and selection criteria. * GENR UHATB=CAL-YHATB GENR MB=(100*ABS(UHATB)/CAL) STAT MB/MEAN=MAPEB GENR UHATB2=UHATB*UHATB GEN1 SSEB=SUM(UHATB2) * * Model C - Cubic Model. * SAMPLE 1960. 1989. AUTO CAL T TSQ T3 / MAX DROP FC / MAX PREDICT=YHATC IBLUP BEG=1989. END=1994. * SAMPLE 1990. 1994. OLS CAL YHATC * * Compute prediction error, error sum of squares, and selection criteria. * GENR UHATC=CAL-YHATC GENR MC=(100*ABS(UHATC)/CAL) STAT MC / MEAN=MAPEC GENR UHATC2=UHATC*UHATC GEN1 SSEC=SUM(UHATC2) * * Model D - Linear-log Model. * SAMPLE 1960. 1989. AUTO CAL LTIME / MAX DROP FC / MAX PREDICT=YHATD IBLUP BEG=1989. END=1994. * SAMPLE 1990. 1994. OLS CAL YHATD * * Compute prediction error, error sum of squares, and selection criteria. * GENR UHATD=CAL-YHATD GENR MD=(100*ABS(UHATD)/CAL) STAT MD / MEAN=MAPED GENR UHATD2=UHATD*UHATD GEN1 SSED=SUM(UHATD2) * * Model E - Reciprocal Model. * SAMPLE 1960. 1989. AUTO CAL INVT / MAX DROP FC / MAX PREDICT=YHATE IBLUP BEG=1989. END=1994. * SAMPLE 1990. 1994. OLS CAL YHATE * * Compute prediction error, error sum of squares, and selection criteria. * GENR UHATE=CAL-YHATE GENR ME=(100*ABS(UHATE)/CAL) STAT ME / MEAN=MAPEE GENR UHATE2=UHATE*UHATE GEN1 SSEE=SUM(UHATE2) * * Model F - Log-linear Model. * SAMPLE 1960. 1989. AUTO LCAL T / MAX DROP FC / MAX PREDICT=YHATF IBLUP BEG=1989. END=1994. * SAMPLE 1960. 1994. GENR SGMASQ=$SSE/$DF GENR YHATF=EXP(YHATF+(SGMASQ/2)) * SAMPLE 1990. 1994. OLS CAL YHATF * * Compute prediction error, error sum of squares, and selection criteria. * GENR UHATF=CAL-YHATF GENR MF=(100*ABS(UHATF)/CAL) STAT MF / MEAN=MAPEF GENR UHATF2=UHATF*UHATF GEN1 SSEF=SUM(UHATF2) * * Model G - Double-log Model. * SAMPLE 1960. 1989. AUTO LCAL LTIME / MAX DROP FC / MAX PREDICT=YHATG IBLUP BEG=1989. END=1994. * SAMPLE 1960. 1994. GENR SGMASQ=$SSE/$DF GENR YHATG=EXP(YHATG+(SGMASQ/2)) * SAMPLE 1990. 1994. OLS CAL YHATG * * Compute prediction error, error sum of squares, and selection criteria. * GENR UHATG=CAL-YHATG GENR MG=(100*ABS(UHATG)/CAL) STAT MG / MEAN=MAPEG GENR UHATG2=UHATG*UHATG GEN1 SSEG=SUM(UHATG2) SAMPLE 1990. 1994. PRINT CAL YHATA YHATB YHATC YHATD PRINT YHATE YHATF YHATG SAMPLE 1994. 1994. PRINT MAPEA MAPEB MAPEC MAPED MAPEE MAPEF MAPEG * DELETE / ALL STOP