Critical assessment of the “internal reference” method to eliminate non-genetic effects within a Combined Family Selection program on the abalone species (Haliotis midae)

Difford, Gareth Frank (2013-12)

Thesis (MScAgric)--Stellenbosch University, 2013.

ENGLISH ABSTRACT: The aim of this study was to critically assess the implementation of the internal reference method within the most recent 173 full-sibling growth trial of the Innovation Fund Abalone Breeding Project. The trial was conducted over two locations for a period of five years, with minimal replication for the majority of test families and a single full-sibling family was entered into each experimental unit (basket) as an internal reference group. The primary focus was firstly, to validate the performance of the internal reference group as a control for comparisons and correction of environmental variation in test family performances. Secondly, to identify areas of weakness and either make recommendations to remedy areas of weakness or justify devoting resources to alternative methods of reducing extraneous environmental variance with limitations on replication. The efficiency and statistical power associated with utilising internal reference information as a covariate and for manual correction respectively were examined for the 6 full-sibling test families that were replicated. This study reports on the evaluation of factors which are potential sources of bias in the internal reference method, the first of which, tag loss, was found to be significant after 6- 12 months. However, it was not found to bias internal reference group performances as the factors which contribute to tag loss were found to act randomly. Variability in size ratio of internal reference to test family at co-stocking proved a significant source of bias, as reference groups smaller than their test family counterparts had reduced performances. Testing for genotype by environment interactions was precluded due to the inherent lack of replication and the subsequent confounding of genotype effects with inter-rearing structure effects at one of the locations. However, significant differences were detected for both traits of interest of the internal reference group over the two locations. Significant antagonistic interactions were detected and identified as a source of bias for average daily weight gain of replicate test families. The evaluation of average daily length gain for the efficiency of adjustment when the internal reference is a covariate and the change in statistical power when the internal reference is used for a manual correction, yielded conflicting results. The latter shows a decrease in statistical power and the former shows an increase in efficiency, both resulting in poor goodness of fit in the respective models. There was however evidence that when no antagonistic interactions occurred “between replicate variance” decreased and therefore the internal reference method has statistical merit provided all critical success factors are satisfied. Recommendations were made for future implementation of the internal reference method to facilitate adequate statistical testing for sources of bias and the prevention thereof. Additionally, an alternative method which may have merit in decreasing environmental variance and the need for replication, is discussed.

AFRIKAANSE OPSOMMING: Die doel van die studie was om die gebruik van ʼn interne verwysingsgroep te ontleed, soos toegepas tydens die evaluering van 173 volsib families as deel van die Innovasiefonds Perlemoen Teelprogram. Die evaluering is gedoen op twee lokaliteite oor 'n tydperk van vyf jaar, met minimale replikasie van die toets families en die gebruik van ‘n enkele volsib familie as 'n interne verwysingsgroep in elke eksperimentele eenheid (mandjie). Die primêre fokus was eerstens om die gebruik van die interne verwysingsgroep vir die korreksie van omgewingsvariasie in die toets familie optredes te evalueer. Tweedens, om spesifieke gebreke te identifiseer ten opsigte van die gebruik van die interne verwysingsgroep en aanbevelings maak dit reg te stel en om die meriete van alternatiewe metodes te oorweeg. Die doeltreffendheid en statistiese onderskeidingsvermoë van die gebruik van interne verwysingsgroep as 'n kovariaat is ondersoek met betrekking tot die 6 volsib groepe wat oor voldoende replikasies beskik het. Die studie doen voorts verslag oor die evaluering van potensiële oorsake van sydigheid ten opsigte van die gebruik van die interne verwysingsgroep, insluitend die beduidende verlies van identifikasie vanaf 6 tot 12 maande. Geen aanduiding van sydigheid is egter gevind en die aanleidende oorsake van verlies van identifikasie blyk van ʼn ewekansige aard te wees. Verskille in die grootte tussen die interne verwysingsgroep en toets-families met aanvang van evaluering blyk 'n belangrike bron van sydigheid te wees, waar die kleiner groepering aan verminderde prestasie gekoppel word. Bepaling van genotipe-omgewing-interaksies kon nie uitgevoer word nie as gevolg van die inherente gebrek van replisering oor lokaliteite. Beduidende verskille is egter waargeneem tussen interne verwysingsgroepe oor die twee lokasies ten opsigte van die beide groei eienskappe. Beduidende antagonistiese interaksies is waargeneem en geïdentifiseer as 'n bron van sydigheid ten opsigte van die gemiddelde daaglikse gewigstoename van replikaat toetsfamilies. Die evaluering van gemiddelde daaglikse lengtetoename met die interne verwysingsgroep as is 'n kovariaat en die verandering in statistiese ontledingsvermoë tydens die gebruik van die interne verwysingsgroep het teenstrydige resultate opgelewer. Laasgenoemde toon 'n afname in statistiese ontledingsvermoë en die eersgenoemde toon 'n toename in doeltreffendheid, met beide swak passing op die onderskeie modelle. In die afwesigheid van antagonistiese interaksies tussen replikasies het variansie afgeneem en beskik die interne verwysingsgroep oor die nodige statistiese meriete indien daar aan al die kritiese vereistes voldoen word. Aanbevelings is gemaak ten opsigte van die toekomstige implementering van die interne verwysingsmetode met verwysing na voldoende statistiese toetsing vir bronne van sydigheid en die voorkoming daarvan. 'n Verdere metode wat oor die nodige meriete beskik om die omgewingsvariasie en die noodsaaklikheid vir replikasie te verminder, word bespreek.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/85658
This item appears in the following collections: