Vol.1, No 10, 2002 pp. 25 - 29
UDC 519.233.5
TESTING IN MULTIPLE REGRESSION ANALYSIS
Vera Djordjević
Faculty of Economics, University of Niš, Serbia and Montenegro
Multiple regression analysis takes great place in the statistical science
and practice. F. Galton, English statistician and geneticist, has written
that the aim of statistical science is "to find out methods in collecting
large groups of data in short, condensend terms convenient for discussion".
This thought, although it was told more than one century ago, can be applied
to the subject and method of multiple linear regression. The significance
of this method is in its usage to predict and estimate the value of this
phenomenon (it has been identified as dependent variable). We have to test
the significance of given results if we want to use the statistics of the
F-test and statistics of the Student's t-test.
Key words: testing, multiple linear regression, regression model,
residual, regression coefficient, standard error of regression, coefficient
of multiple determination, zero hypothesis, alternative hypothesis, analysis
of variance, F-test, t-test.
TESTIRANJE MODELA LINEARNE VIŠESTRUKE
REGRESIJE
Upotrebljivost modela linearne višestruke regresije za predviđanje i ocenjivanje
vrednosti zavisne promenljive proverava se testiranjem. Za testiranje možemo
koristiti F-test i t-test. I jednim i drugim testom proverava se da li
oblik zavisnosti utvrđen na osnovu raspoloživih podataka (podataka uzorka)
važi za osnovni skup. Samo u tom slučaju regresioni model može poslužiti
analitičaru za dalju statističku analizu i korišćenje.
