$BKY!w9a@nBg3X7P:Q3XIt$G$9!#(B $B2,K\$5$s!w6bBtBg3X(B $B$O=q$-$^$7$?!'(B[fpr 568] >$B!!6h4V?dDj$r9-$a$k$3$H$bI,MW$@$H;W$$$^$9!#(B $B$3$C$A$b=EMW$G$9$M!#(B >$B!!M-0U:9$,$"$l$P!"8!DjNO1>!9$OLdBj$K$5$l$J$$$G$7$g$&!#(B $B8z2LNL(Beffect size $B$N>l9g$O!"M-0U:9$,$"$C$F$b8z2LNL$rLdBj$K$7(B $B$^$9!#(B Keppel,G.(1991) Design and analysis. 3rd ed. Prentice-Hall. p87 $BA0$K0zMQ$7$?D>A0$G$9!#(Bomega squared $B&X(B2 $B$,$3$3$G$O8z2LNL$G$9(B $B!#(B Since estimated omega squared is essentially uninfluenced by sample size, its magnitude will provide useful information regardless of the underlying power of the experiment on which it based. It is for this reason that editors and reviewer urge that we estimate relative treatment magnitude for all statistical tests we perform. A "small" but significant F might suggest the presence of a "trivial" effect that was detected by a particularly powerful experiment, whereas a "medium" but nonsignificant F might suggest the possible presence of an "important" effect that was not detected because of a serious lack of power. $B$=$NA0$+$iFI$a$P!"
l9g$H$o$+$j$^$9$,!"M-0U(B $B:9$,$"$C$F$b:3:Y$J:9$G$"$k$+>.$5$J:9$+CfDxEY$N:9$+Bg$-$J:9$+(B $B$J$I$r9M$($^$9!#(B $B9a@nBg3X7P:Q3XIt(B $B!!!!!!!!!!(B $B!!KY!!7http://fourier.ec.kagawa-u.ac.jp/~hori/