Nonlinear models applied to seed germination of Rhipsalis cereuscula Haw (Cactaceae)

The objective of this analysis was to fit germination data of Rhipsalis cereuscula Haw seeds to the Weibull model with three parameters using Frequentist and Bayesian methods. Five parameterizations were compared using the Bayesian analysis to fit a prior distribution. The parameter estimates from t...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Acta scientiarum. Technology 2014-01, Vol.36 (4), p.651-656
Hauptverfasser: Guedes, Terezinha Aparecida, Rossi, Robson Marcelo, Martins, Ana Beatriz Tozzo, Janeiro, Vanderly, Carneiro, José W P
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The objective of this analysis was to fit germination data of Rhipsalis cereuscula Haw seeds to the Weibull model with three parameters using Frequentist and Bayesian methods. Five parameterizations were compared using the Bayesian analysis to fit a prior distribution. The parameter estimates from the Frequentist method were similar to the Bayesian responses considering the following non-informative a priori distribution for the parameter vectors: gamma ([10.sup.3], [10.sup.3]) in the model [M.sub.1], normal (0, [10.sup.6]) in the model [M.sub.2], uniform (0, [L.sub.sup]) in the model [M.sub.3], exp ([mu]) in the model [M.sub.4] and [L.sub.normal] (p, [10.sup.6]) in the model [M.sub.5] . However, to achieve the convergence in the models [M.sub.4] and [M.sub.5], we applied the u from the estimates of the Frequentist approach. The best models fitted by the Bayesian method were the [M.sub.1] and [M.sub.3]. The adequacy of these models was based on the advantages over the Frequentist method such as the reduced computational efforts and the possibility of comparison. Keywords: Bayesian inference, growth curve, modeling. Neste estudo, foi proposto o ajuste de germinacao de sementes pelo modelo Weibull com tres parametros por meio da metodologia frequentista e da Bayesiana. Na analise Bayesiana foram utilizadas cinco parametrizacoes para as distribuicoes a prior nao-informativas e foram comparadas quanto ao ajuste. As estimativas dos parametros obtidas pela metodologia frequentista foram similares aos da metodologia Bayesiana quando considerado distribuicoes a priori nao-informativas para o vetor de parametros: gama (103, 103) no modelo [M.sub.1], normal (0, 106) no modelo [M.sub.2], uniforme (0, [L.sub.sup]) no modelo [M.sub.3], exp (u) no modelo [M.sub.4] e lognormal (p, 106) no modelo [M.sub.5]. No entanto, para a convergencia nos modelos [M.sub.4] e [M.sub.5], foi utilizado para u os valores obtidos pela metodologia frequentista. Os melhores modelos para a modelagem Bayesiana foram os modelos [M.sub.1] e [M.sub.3]. Estes modelos foram considerados adequados, tendo como vantagem sobre a metodologia frequentista o menor esforco computacional e a possibilidade de comparacao. Palavras-chaves: inferencia Bayesiana, curva de crescimento, modelagem.
ISSN:1806-2563
1807-8664
DOI:10.4025/actascitechnol.v36i4.21192