9429 - Laserfiche WebLink

Home Browse Search

J. Wildl.:~tanage. 63(3):1999 $TdTISTICAL $ICNiFIC.4NCE TESTING • )vJutson 169 WHAT ARE THE ALTERNATIVES? What should we do instead of testing hypoth- eses? As Quinn and Dunham (1983) pointed out; it is more fruitful to determine the relative importance to the contributions of, and inter- actions between, a number of processes. For this purpose, estimation is far more appropriate than hypothesis testing (Campbell 1992). For certain other situations, decision theory is an appropriate tool.. For either of these applica- tions, as well as for hypothesis testing itself, the Bayesian approach offers some distinct advan- tages over the traditional methods. These alter- natives are briefly outlined below. Although the alternatives will not meet all potential needs, they do offer attractive choices in many fre- quently encountered situations. Estimates and Confidence Intervals Four decades ago, ?,nscombe (1956) ob- served that statistical hypothesis tests were to- tally irrelevant. and that what was needed were estimates of magnitudes of effects, with stan- dard errors. Yates ;196.1) indicated that "The most commonly occurring weakness in the. ap- plication of Fisherian methods is undue em- phasis on tests of significance. and failure to reco~rtize that itt many hies of experimental work estimates of the treahnent effects, togeth- er ~~Zth estimates of the errors to ~.vhich then are subject, are the quantities of pntnan• interest." Further. he(:ause ~.il(ilile ecolo~~ists want to in- fluence management practices. Johnson (19951 noted that. "If ecologists are to he taken seri- o!_rsl~• b~- decision makers. they unist provide in- formation useful li)r decidincg on a course of ac- tion, as opposed to addressirng pccrc-l~~ academic c{nestions.•' To enlince that point. severil echi- cation and psvcholo~~ical jountals hove adopted editorial policies requiriu~ that parameter esti- mates accontpanv any P-values he presented !~(cLean and 1;ntest 1~)~)tii, Ordiuarv cunfidenc•e intervals pro~~de more inli)rmation than clo P-values. Knowing that a 95°!n confidence inten~al includes zero tells one that, if a test of the lhypothesis that the param- eter equals zero is antchtcted, the resulting P- value ~~ill be >0.0.5. A confidence interval pro- vides both an estimate of the effect size and a me.tsure of its uncertainh~. A 9S°7o confidence interval of. sa., (-50.:300) suggests the param- eter is less well estimated than would a confi- dence inten-al of (L0. 1:30). Perhaps surpris- ingly, confidence intervals have a longer history than statistical hypothesis tests (Schmidt and Hunter 1997). With its advantages and longer history, ~vhy have confidence intervals not been used more than they have? Steiger and Fouladi •(1997) and Reichardt and Gollob (1997) posited several ex- planations: (1) hypothesis testing has become a tradition; (2) the advantages of confidence in- tervals aze not recognized; (3) there is some ig- norance of the procedures available; (4) major. statistical packages do not include many confi- dence interval estimates; (5) sizes of parameter estimates are often disappointingly small, even though they may be very significantly different from zero; (6) the wide confidence intervals that often result from a study are embarrassing; (7) some hypothesis tests (e.g., chi-squaze contin- gency=table) have no uniquelc defined param- eter associated with them: and (8i recommen- dations to use confidence intervals often are ac- companied b~~ recommendations to abandon .statistical tests altogether. which is umvelcome advice. These reasons are not valid excuses for avoiding confidence inten•ais in lieu of h~poth- esis tests in situations for ~~•hich parameter es- tintation is the objective. Decision Theory Otten exTeriments or surveys are conducted to help ntal:e some decision, suc!t BLS ~vlhat limits to set on huntin<_ se.LSOUS. If a forest stare! shoctkl he lc>~,ecl. or i} :t pesticicle slu)uld be appro~~ecl. In those r.LSes. ll~rotltesis testing is inadequate. for it does not hue utto consideration tte costs o ~dtentative actions. Here a nsefitl trxtl is statistical decision tteon~: the theon~ of actin, rttionallv .~itc respect to anticipated ~,ains and losses. in the lacy of nncertainh. Ehpottesis testincg ,euerallv liu!its the prohabilih~ of a T.pe I error (rejecting a tnte null h~jtottesis i, oltett arhitrarily set ut a . = 0.0.5. while letting tte prohahilih~ of a Tire II error tacce_ptincg a false n!ill It~pothesis~ fall where it may. In ecologicil situations. however. a Type II error tnav be far more costs than a Type I error (Tott and Shea 1983)..•~s art example, ap- pro~in~g a pesticide that reduces the survih~il rate of an endan_ered species b~' S9o may be disas- trous tci that species, even if that change is not statistically detectable. As another, continued overharvest in marine fisheries may resrilt in the collapse of tte ecosystem even while statistical tests are unahle to reject the mill hypothesis that fishing has no effect (Dayton 1998). Details on