Lecture 2/ Lecture 3

Question 1

why do we use linear regression in twin modeling?

what are paralelles?

Answer 1

Twin design: decompose phenotypic variance into genetic and environmental components

linear regression: decompose variance of a dependent variable into an explained variance component and unexplained residual component

Question 2

what is probably the case if

1. rMZ > rDZ
2. rMZ ~ rDZ
3. rMZ < 1

Answer 2

1. genetic factors play a role
2. environmental factors play a role, genetics not so much
3. non shared environmental factors are important

> this is always the case

> residual variance is included in E!

Question 3

falconer: what is h²?

what is the difference between h² and a²?

Answer 3

h²: proportion of variance explained by genetic differences

> a² is h² standardized

> if var(pheno) = 1

then

h² == a²

Question 4

v.Beek paper (alcohol abuse)

what was aim of the study?

what was the conclusion?

Answer 4

aim:

> investige how genetic fisk factors for symptoms of alcohol abuse and dependance change across age

conclusion:

1. genetic influence increased with age
2. shared environmental influences decreased with age
3. unique environmental influence increased with age

>>> same genetic factors influences AAD at all ages, but influence of this factor changes with age

Question 5

why is model fitting better than just using falconers formula?

Answer 5

proper model fitting provides insight into variability of parameter estimates!

Question 6

model fitting: differences between saturated and ACE model?

Answer 6

saturated model: uses observed covariance matrix

ACE model: uses estimated covariace matrix

>>> compare those two estimate fit of the ACE model

Question 7

the saturated is not really….?

why use it anyways?

Answer 7

saturated model it technically no model, but

1. it gives you the most precise observed statistics
2. you can test assumptions with it about, mean/variance differences or formally test differences in correlations between zygosity
3. it provides a baseline for testing other models

Question 8

nested models: which one most complex?

which one most constrained?

which model has the best fit?

why not always use this model?

Answer 8

most complex: saturated model

most constrained: E model (everything dropped and equal but E)

>>> most complex models with most free parameters always has the best fit

>>> try to use a parsimonious model, so find a model that is most contrained without using significant amount of fit

Question 9

what are 3 general laws of twin modeling?

Answer 9

1. all traits show significant genetic influence (h² > 0)
2. no traits are 100% heritable (h² < 0)
3. heritability is caused by many genes of small effect

Question 10

plomin & deary:

> 5 findings on IQ

Answer 10

plomin & deary

1. heritability increases dramatically with age despite genetic stability
2. intelligence indexes genera genetic effects across diverse cgnitive and learning abilities
3. assortative mating is greater for intelligence than for other traits
4. genetic causes of high inteligence are quantitatively, not qualitatively different from the rest of the distribution ( low IQ however can be qualitatively different)
5. intelligence is correlated with alot of other stuff such als social mobility, health, illness and mortality

Question 11

what is a quantitiative difference in genes?

what is qualitative?

Answer 11

quantitative: differences arising through differences at many different loci ( high IQ)
qualitative: difference arising through one genetic difference e.g. trisomy 21 > down syndrome (low IQ)