摘要
Inthisstudy,thesampledatawasbasedon2790samplesof9treesfromdahurianlarch(Larixgmelinii)plantationslocatedinQitaiheForestBureauinHeilongjiangProvince.ThemodifiedLogisticmodely=b1/[1+exp(b2x)]+b3wasselectedtomodelingmicrofibrilanglefromsixmodelsbasedonnonlinearregression.Then,thelogisticmodelwasfittedusingnonlinearmixed-effectmodelingapproachbasedonNLMEofS-PLUSsoftware.Evaluationstatistics,suchasAIC,BIC,LogLikelihoodandLikelihoodratiotestwereusedformodelcomparisons.TheresultsshowedthattheLogisticmodelwithparametersb1,b2,b3asmixedeffectsshowedthebestperformance.Correlationstructuresincludedcompound-symmetrystructure(CS),first-orderautoregressivecorrelationstructureAR(1),movingaveragecorrelationstructureMA(1)andautoregressive-movingaveragecorrelationstructure[ARMA(1,1)]wereincorpratedintothebestmicrofibrilanglemixedmodel.[ARMA(1,1)]significantlyimprovedtheprecisionofmixedmodel.Validationconfirmedthatthemixedmodelwithcalibrationofrandomparameterscouldprovidemoreaccurateandpreciseprediction.Therefore,theapplicationofmixedmodelnotonlyshowedthemeantrendsofmicrofibrilangle,butalsoshowedtheindividualdifferencebasedonvariance-covariancestructureandcorrelationstructure.
出版日期
2012年03月13日(中国期刊网平台首次上网日期,不代表论文的发表时间)