Pesquimat (Sep 2014)

IDENTIFICACIÓN DE OBSERVACIONES INFLUYENTES EN LA DISCRIMINACIÓN DE MUESTRAS DEL GÉNERO MINTHOSTACHYS DE CAJATAMBO Y UNCHOS

  • Doris Gómez Ticerán,
  • Olga Solano Dávila,
  • Víctor Osorio Vidal,
  • Liliana Huamán del Pino,
  • Gregoria Natividad Ramón Quispe,
  • Jorge Ícaro Condado Jáuregui,
  • María Enriqueta Gallardo Cabanillas,
  • Yakov Mario Quinteros Gómez

DOI
https://doi.org/10.15381/pes.v15i2.9609
Journal volume & issue
Vol. 15, no. 2

Abstract

Read online

This paper explores the possibility of identifying influential observations in discriminant analysis framework, 160 botanical specimens of the genus Minthostachys, pubescent and pubescent not collected in the province of Cajatambo department of Lima and Unchos departarment of Ancash. The evaluations of morphological variables in tho main branch of each Minthostachys being studied were: length of petiole, leaf length and width of the blade. Taxonomic and systematic studies of the samples were performed at the Laboratory of Ethnobotany and Economic Botany of the Natural History Museum and the determination of the species are held in the herbarium of the San Marcos University, using the Cronquist classification system, for Cajatambo which marked 51 plants such as non-pubescent and pubescent 49: while for Unchos ranked 40 with abundant pubescence band 20 escaca pubescence. For the 160 Minthostachys plants, eliminating one at a time, we calculated the value of the Mahalanobis distance, the probability of misclassification and the scores of the Fisher discriminant function (Campbell, 1978: Fung, 1992. 1995). The analysis correctly discriminated 151 plants Minthostachys ie, 94.4 % of a total of 160, a value largo enough to show the effectiveness of discriminant analysis. For the full sample and removing each time one of the samples or observations, we calculated the value o the Mahalanobis Distance, the probability of misclassification, the weightings and scores of discriminant function of Fisher. Comparison of the values of the estimates, with and without the observation under evaluation, it was concluded that observations 110, 112 and 114 were identified as influential.

Keywords