Sveučilište u Zagrebu Agronomski fakultet

Obrana doktorskog rada :: Mr.sc. Borut Bosančić, dipl.ing.agr.

Datum održavanja:

Application of Meta-Analysis Techniques in Horticultural Research: Simulated and Real Research Data
petak, 27.04.2018. :: 12:00 sati :: Velika vijećnica

 Povjerenstvo za obranu doktorskog rada:

  • prof.dr.sc. Jerko Gunjača, Sveučilište u Zagrebu Agronomski fakultet,
  • prof.dr.sc. Anamarija Jazbec, Sveučilište u Zagrebu Šumarski fakultet,
  • prof.dr.sc. Nikola Mićić, Sveučilište u Banja Luci Poljoprivredni fakultet.

Mentorica:

  • prof.dr.sc. Marija Pecina, Sveučilište u Zagrebu Agronomski fakultet.

Summary

In numerous horticultural researches the authors’ interest is related to changes in fruit characteristics caused by treatment in one or more genotypes. However, the results are often contradictory. The general significance and the size of the treatment effect can be estimated reliably by meta-analysis (MA) using metadata.

In horticultural research there are no known developed procedures for the use of meta-analysis which represented a challenge to generate it in order to process the typical multilayer data profile originating from multiple sources and multiple measurements scales. The goal of this research was to adapt and implement meta-analysis on studies typical for horticultural science.

This was done through both simulations and analysis of a typical real horticultural data - in apple as a model crop. In the databases the search for adequate studies was conducted and chosen were those ones exploring the effect of hail nets on fruit characteristics. In order to analyse and present the results in the best possible manner, a novel Main Effects Meta Principal Components Analysis (ME Meta-PCA) was developed. Recommendations are provided on how to conducts, use, report and interpret MA.

Upon analysis of the typical agronomical experimental dataset simultaneously by meta-analysis and the analysis of variance, the reliability of metadata analysis by MA in horticulture was effectively verified. The advantage of MA is direct and reliable, yet comprehendible presentation of the individual study level and overall effect size in form of forest plot. Practical analysis of hail netting effect in apple from the simple to most complex multivariate multiscale setting across multiple studies for multiple cultivars was conducted. Apple was used as a model crop to provide wide variation with its many varieties, and therefore the results are easily applicable for other horticultural crops. The meta-analysis provided valuable results by revealing general patterns on fruit characteristics and cultivars response under hail netting.

Use of meta-analysis in horticulture is therefore recommended as obligatory in future studies and systematic reviews where the research goals require comprehensive analysis and general overview of treatment effects.

Key words: Metadata, main effects, PCA, treatment effect, biometrics