Procjena genetske korelacije između binarne i normalne varijable koristeći GLIMMIX SAS proceduru

master's thesis

• Title (english): Evalutation of genetic correlation between binary and normal variables using the GLIMMIX SAS procedure
• Author: Mateja Ćosić
• Mentor: Maja Ferenčaković (mentor)
• Committee member: Maja Ferenčaković (predsjednik povjerenstva)
• Committee member: Vlatka Čubrić Čurik (član povjerenstva)
• Committee member: Ino Čurik (član povjerenstva)
• Granter: University of Zagreb; Faculty of Agriculture; (Animal Science); Zagreb
• Defense date and country: 2019-09-30, Croatia
• Scientific / art field, discipline and subdiscipline: BIOTECHNICAL SCIENCES; Agronomy

Abstract

Procjena genetske korelacije između dvije varijable koje imaju normalnu raspodjelu je
uobičajena i može se napraviti bez većih poteškoća, dok se problem javlja kad varijable nemaju
normalnu raspodjelu, kao što je binarna varijabla. U tom slučaju potrebno je koristiti
generalizirani bivarijatni mješoviti model, a jedna od opcija je korištenje GLIMMIX procedure.
GLIMMIX procedura je SAS procedura koja prilagođava modele podatcima koji nemaju
normalnu raspodjelu. U radu su simulirani podatci uspješnosti koncepcije i težine kod starosti
od 12 mjeseci 5000 krava, kćeri 50 bikova. Provedena je bivarijatna analiza između binarne
(uspješnost koncepcije) i normalne (težina) varijable koristeći GLIMMIX proceduru. Radi
usporedbe provedena je bivarijatna analiza tretirajući uspješnost koncepcije kao normalnu
varijablu, te univarijatne analize za obje varijable. Kod analiza binarnih podataka korištene su
logit i probit funkcije veze. Iz dobivenih rezultata analiza zaključeno je da tretiranjem binarne
varijable kao normalne, ne dobivamo precizne procjene varijanci i kovarijanci, a time ni
genetske korelacije. Stoga GLIMMIX procedura znatno olakšava ovakve složenije analize
proširenjem mogućnosti generaliziranih mješovitih modela.

Abstract (english)

Genetic correlation estimation between two variables that have a normal distribution is
common and can be done without much difficulty. The problem arises when one of the
variables does not have a normal distribution, i.e. binary variable. In this case, a generalized
bivariate mixed model should be used and GLIMMIX procedure can help here. The GLIMMIX
procedure fits statistical models to data where the response is not necessarily normally
distributed. In this master’s thesis, success of conception and weight of 5 000 cows that are
12 months old, daughters of 50 bulls, were simulated. A bivariate analysis was performed
between binary (success of conception) and normal (weight) variables using GLIMMIX
procedure. For comparison, bivariate analysis was performed treating success of conception
as normal variable and univariate analysis for both variables. Logit and probit link functions
were used for binary data analyzes. The results showed that, by treating binary variable as
normal, did not lead to accurate estimates of variance and covariance, and thus genetic
correlation. Therefore, the GLIMMIX procedure noticeably facilitates more complex analyzes
by extending the abilities of generalized mixed models.

• Keywords: bivarijatni model; binarna varijabla; normalna varijabla; GLIMMIX procedura
• Keywords (english): bivariate model; binary variable; normal variable; GLIMMIX precedure
• Language: croatian
• URN:NBN: urn:nbn:hr:204:288388
• Study programme: Title: Animal Genetics and Breeding; Study programme type: university; Study level: graduate; Academic / professional title: magistar/ magistra inženjer/inženjerka genetike i oplemenjivanja životinja (magistar/ magistra inženjer/inženjerka genetike i oplemenjivanja životinja)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Access restricted to students and staff of home institution
• Created on: 2019-11-18 07:48:12