发帖
雷达卡
Problem(Abstract)
[size=1em]I would like to analyze cross-sectional time series (panel) data using SPSS Statistics. The SPSS Trends package only allows you to model one series at a time. Is there a way to model cross-sectional time series in SPSS Statistics?
Resolving the problem
[size=1em]Some cross-sectional time series may be analyzed using mixed linear modeling procedures. In this solution, we provide an example of this kind of model using the MIXED procedure SPSS Statistics. The GENLIN procedure, which offers GEE (generalized estimating equations) estimation is also available. Beginning in Release 19, the GENLINMIXED procedure is available for fitting generalized linear mixed models.
The example we provide reproduces the results for a Fuller-Battese model (Fuller and Battese, 1974) presented in Littell, Milliken, Stroup, and Wolfinger (1996; pp. 130-134). Here is a brief description of the data for this example.
The dependent variable D measures per capita demand deposits. D is measured for 7 states over 11 years. There are four concurrently measured independent variables:
Y permanent per capita personal income
RD service charge on demand deposits
RT interest on time deposits
RS interest on savings and loan association share.
As Littell et al. (1996) explain, the underlying econometric model is multiplicative, but it can be transformed into a standard linear model by taking the natural logarithms of the variables and using the log-transformed variables in a mixed linear modeling procedure.
The following command syntax reads in the data and performs the necessary log transformations.
_______________
DATA LIST FREE
/ state (A2) year d y rd rt rs .
BEGIN DATA
CA 1949 533 1347 0.343 1.114 2.905
CA 1950 603 1464 0.364 1.162 2.935
CA 1951 669 1608 0.367 1.493 3.093
CA 1952 651 1636 0.369 1.567 3.073
CA 1953 609 1669 0.410 1.594 3.357
CA 1954 634 1716 0.499 1.609 3.295
CA 1955 665 1779 0.496 1.637 3.451
CA 1956 676 1878 0.533 1.757 3.539
CA 1957 642 1963 0.630 2.641 3.930
CA 1958 678 2034 0.667 2.641 3.982
CA 1959 714 2164 0.664 2.648 4.047
DC 1949 854 1603 0.261 0.676 2.803
DC 1950 1013 1773 0.267 0.662 2.877
DC 1951 1185 2017 0.266 0.677 3.006
DC 1952 1076 1921 0.267 0.729 2.975
DC 1953 1004 1856 0.287 0.883 3.035
DC 1954 1044 1868 0.308 1.500 3.083
DC 1955 1067 1931 0.318 1.504 3.177
DC 1956 1062 1951 0.322 1.598 3.250
DC 1957 1120 2085 0.346 2.231 3.368
DC 1958 1196 2144 0.360 2.100 3.457
DC 1959 1168 2167 0.418 2.342 3.727
FL 1949 408 1024 0.354 0.909 2.314
FL 1950 433 1007 0.342 0.957 2.327
FL 1951 469 1068 0.335 1.002 2.428
FL 1952 470 1068 0.328 1.052 2.577
FL 1953 464 1138 0.354 1.118 2.625
FL 1954 465 1137 0.374 1.268 2.871
FL 1955 545 1306 0.378 1.339 2.882
FL 1956 567 1339 0.399 1.486 3.032
FL 1957 531 1383 0.447 2.420 3.338
FL 1958 533 1409 0.498 2.453 3.353
FL 1959 522 1457 0.523 2.489 3.575
IL 1949 843 1465 0.143 0.852 2.504
IL 1950 860 1468 0.146 0.847 2.448
IL 1951 887 1555 0.147 0.936 2.449
IL 1952 914 1648 0.144 1.059 2.568
IL 1953 909 1711 0.150 1.091 2.703
IL 1954 928 1775 0.164 1.130 2.748
IL 1955 939 1815 0.172 1.141 2.778
IL 1956 944 1915 0.183 1.354 2.932
IL 1957 899 1980 0.203 1.628 3.155
IL 1958 919 2001 0.214 1.737 3.402
IL 1959 874 2035 0.231 2.054 3.497
NY 1949 1370 1492 0.112 0.687 2.099
NY 1950 1405 1515 0.119 0.724 2.082
NY 1951 1409 1566 0.119 0.795 2.218
NY 1952 1421 1659 0.120 1.050 2.435
NY 1953 1395 1744 0.134 1.241 2.477
NY 1954 1415 1802 0.145 1.346 2.540
NY 1955 1431 1808 0.146 1.406 2.655
NY 1956 1416 1916 0.168 1.754 2.774
NY 1957 1443 2074 0.189 2.231 2.957
NY 1958 1453 2120 0.192 2.360 3.073
NY 1959 1417 2197 0.203 2.521 3.223
TX 1949 573 995 0.149 0.839 2.755
TX 1950 634 1052 0.147 0.836 2.740
TX 1951 679 1154 0.148 0.812 2.819
TX 1952 668 1176 0.147 1.070 2.880
TX 1953 666 1228 0.160 1.170 3.082
TX 1954 708 1285 0.182 1.328 3.093
TX 1955 722 1335 0.191 1.368 3.071
TX 1956 708 1358 0.208 1.544 3.068
TX 1957 675 1416 0.250 2.121 3.487
TX 1958 716 1457 0.278 2.241 3.413
TX 1959 703 1520 0.303 2.435 3.671
WA 1949 418 1146 0.358 0.937 2.068
WA 1950 501 1324 0.361 0.973 2.229
WA 1951 525 1433 0.365 1.039 2.367
WA 1952 519 1481 0.381 1.305 2.553
WA 1953 500 1531 0.414 1.342 2.848
WA 1954 537 1602 0.481 1.348 2.865
WA 1955 545 1649 0.529 1.770 2.907
WA 1956 525 1656 0.587 1.779 3.011
WA 1957 494 1711 0.681 2.313 3.252
WA 1958 521 1754 0.716 2.302 3.306
WA 1959 515 1809 0.730 2.495 3.507
END DATA.
COMPUTE logd = LN(d).
COMPUTE logy = LN(y).
COMPUTE logrd = LN(rd).
COMPUTE logrt = LN(rt).
COMPUTE logrs = LN(rs).
_______________
The command syntax below reproduces the inferential statistics for the model provided on pp. 132-133 of the Littell et al. (1996) reference.
_______________
MIXED
logd BY state year WITH logy logrd logrt logrs
/CRITERIA = CIN(95) MXITER(100) MXSTEP(5) SCORING(1)
SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE)
LCONVERGE(0, ABSOLUTE)
PCONVERGE(0.000001, ABSOLUTE)
/FIXED = logy logrd logrt logrs | SSTYPE(3)
/METHOD = REML
/PRINT = SOLUTION
/RANDOM = state year | COVTYPE(VC) .
_______________
References
Fuller, W.A., & Battese, G.E. (1974). Estimation of linear models with crossed error structure. Journal of Econometrics, 2, 67-68.
Littell, Ramon C., Milliken, George A., Stroup, Walter W., & Wolfinger, Russell D. (1996). SAS System for Mixed Models. Cary, NC: SAS Institute
Historical Number
[size=1em]25097
京公网安备 11010802022788号
