Modelling

Question 1

mulitlevel modelling operates between 2 extremes

Answer 1

data are so highly correlated, n=1 per unit and data are completely uncorrelated
works by estimating which data needs ‘pooling’ or shrinkage - so reduces df to 1 for each unit

Question 2

what are summary measures not good with?

Answer 2

unbalanced design

Question 3

what is a binomial distribution?

Answer 3

two possible outcomes are equally likely

Question 4

what is a binary response?

Answer 4

Y/N, absent/present

Question 5

what does a general linear model assume about dist?

Answer 5

assumes that unexplained variation is normally distributed

Question 6

what does generalised linear model assume about dist?

Answer 6

assumes that unexplained variation can follow some other known distribution

Question 7

types of distributions and what they are

Answer 7

log- normal = if effects are multiplicative not additive
exponential = eg latency/survival, probability of evrything remains constant, waiting for event
weiball = survival with non-constant mortality
poission = random rare, discrete events
negative binomial= clustered discrete events

Question 8

when is bootstrapping used?

Answer 8

when you want to estimate parameter of population and want to get estimate of CI

Question 9

when is bootstrapping used? and what sample size?

Answer 9

when you want to estimate parameter of population and want to get estimate of CI for sample size larger than 50

Question 10

what are contrasts?

Answer 10

allow for testing of pair-wise differences after ANOVA

Question 11

forward multiple regression

Answer 11

start with no varialbes then add most sig and then next most sig

Question 12

backwards multiple regression

Answer 12

start with all variables remove least sig, then so on

Question 13

Stepwise multiple regression

Answer 13

start the same as forwards with no variables but at any time can remove non-sig terms

Question 14

What measure do we use for the balance between fit of a model and no of parameters it measures?

Answer 14

Akaike information crieterion (AIK)

Question 15

what is logistic regression? and what statistic does it use?

Answer 15

Characterised by a link response distribution (binomial) and a link function which transforms mean value to make it more linear

Uses z statisitc

Question 16

what is deviance?

Answer 16

Deviance is a measure of goodness of fit of a generalized linear model.

Question 17

Deviance in logistic regression should not be

Answer 17

residual deviance should not be 2x as large as df

Question 18

what is a GLM?

Answer 18

The general linear model incorporates a number of different statistical models: ANOVA, ANCOVA, MANOVA, MANCOVA, ordinary linear regression, t-test and F-test. The general linear model is a generalization of multiple linear regression model to the case of multiple predictors
GLM: error is Normal (mean = 0, sd = )

Question 19

3 examples of a generalised linear model

Answer 19

linear regression, logistic regression and Poisson regression.
Generalized LM: error is… lots of possibilities

Question 20

Skewed data because of two many zeros?

Answer 20

zero inflated data

Question 21

what tests can you do when you have outliers?

Answer 21

parametric on ranked data, non parametric or permutation tests

Question 22

what is survival anaylsis?

Answer 22

analysing expected duration of time unitl one or more events happen eg death

Question 23

what is censoring

Answer 23

when the actual data point isnt known but you can set boundaries based on what it must have been

Question 24

brackets in R 
() 
{}
[]
<>

Answer 24

() = using to bound an object during execution of a function
{} = used to bound creation of a function
[] = used to subscript an object
<> = denotes greater or less than

Question 25

what is data mining?

Answer 25

lots of candidate predictors but no strong theory to predict which should be important

Question 26

2 general classes of cluster based anaylsis

Answer 26

supervised learning- know true identity of some clusters and use these to make predictive models when you dont know group memebership unsupervised learning - dont know whats right or wrong so try and find natural clustering patterns in data

Question 27

k-means clustering is an example of ...

Answer 27

unsupervised learning

Question 28

types of supervised clustering

Answer 28

DFA, logistic regression, mulimonial logisitc regression, neural networks, genetic algorithms

Question 29

minkowski distance

Answer 29

a generalisation of a distance measure with eucidean and manhattan as special cases

Question 30

measuring between data points is what type of clustering?

Answer 30

hierachical or agglomerative clustering

Question 31

what is loess?

Answer 31

way to run a smooth average but it weighted

Question 32

piecewiese regressions and splines?

Answer 32

piecewise - IV is partitioned into intervals and a sep line segment is fit to each interval splines - fits polynomial sections then joins them up

Question 33

name all 3 orthoganl contrasts

Answer 33

helmert, difference and polynomial

Question 34

name 2 non orthogonal contrats

Answer 34

treatment and simple

Question 35

how can we reduce influence of outliers?

Answer 35

use robust statistics

Question 36

why is stepwise regression dangerous?

Answer 36

automatic so can miss effects of outliers, non-linearity and non-normality. there is an unseen inflation of false positives and cant use categorical so hve to code as contrasts first

Question 37

3 benefits of multilevel modelling

Answer 37

gain df and power, can deal with unbalanced designs, predictors can be of diff levels and all incorporated into one model

Question 38

4 examples of multivariate approach

Answer 38

PCA, FA, LDA/DFA, MANOVA

Question 39

What is shrinkage? in multimodel

Answer 39

when data values arent any more correlated within levels than between

Question 40

Two cautions of multivariate approaches?

Answer 40

At each level you are estimating variances, so if n is small at any one level, estimates may be wildly out. Methods were developed for BIG datasets Different algorithms used in different packages

Question 41

PCA and its assumptions | what is the PC1 and PC2

Answer 41

PCA can be used for data reduction, it is a simple linear transformation of robust data, so doesn’t depend upon assumptions about the data’s distribution. It is a rotation of the original axes to create new axes such that --The First Principal Component (PC1) is, BY DEFINITION, the single axis that accounts for most of the variation in the original variables. This is the axis along which there is the tightest covariation of the original variables. -- The Second Principal Component (PC2) is, BY DEFINITION, the NEXT single axis that accounts for most of the 2nd greatest amount of variation in the original variables, subject to the constraint that it is at right angles to (‘orthogonal’ or independent’ of) the first axis. Components may be interpretable from their coefficients

Question 42

PCA on raw data is called

Answer 42

covariance

Question 43

PCA on standardised 'z scores'

Answer 43

correlation, z score is mean of 0 and SD of `1 | THIS IS THE NORM

Question 44

what graph for a pca to help interpretation?

Answer 44

biplot

Question 45

how many components is enough in PCA?

Answer 45

Eigenvalue > 1, use screeplot and look for 'natural break' Subjective balance between lots of variation captured’ and ‘not too many components’ - harder to justify

Question 46

Although PCA doesn't rely on normality of variables, the answers you get will be more robust (and sensible) if ...

Answer 46

there are no outliers having a big influence on the correlations between variables, and (ii) the relationships are linear.

Question 47

Factor analysis

Answer 47

Conceptually similar to PCA, but underpinning logic rather different Assumes there is a ‘hidden variable’ that drives the observed variables and their relationships

Question 48

PCA vs FA | 5 points

Answer 48

PCA - total variance FA - shared variance PCA - unaffected by no you chose to work with FA - no of factors changes coeffiecients PCA - rotation of orginial variables FA- creates new axes then rotates these PCA - data reduction FA - not data reduction, uncovering hidden variables PCA - doesnt rely on normality FA - does rely on multivariate normality

Question 49

what is verimax?

Answer 49

factors will have either large or small loadings of any particular variable Aids interpretability. Most popular

Question 50

what does it mean if you get sig p in FA?

Answer 50

There is a highly significant difference between the variance captured by the factor and the variation in the original variables. These factors are not enough.

Question 51

DFA/LDA

Answer 51

Use to find the best (linear) separation between groups, based on multiple dependent (response) variables Useful for generating predictions about group membership of new items (new data

Question 52

what is Wilk's lambda used in?

Answer 52

MANOVA

Question 53

what are canonical variates

Answer 53

The 'best separating dimensions' in MANOVA

Question 54

what is the Reverse of LDA? | and what is an alternative of LDA?

Answer 54

MANOVA - reverse | Logistic reg- alternative

Question 55

Assumptions of LDA/DFA

Answer 55

Sample size of smallest group > number of predictors Best to have at least 4-5 times as many observations as predictors Normality of predictors (outliers are fatal, some skew is OK) Homogeneity of variances & covariances: important (can use z-scores) if some predictors highly correlated (multicolinearity) then analysis may fail or give unreliable results

Question 56

what is use To test for an association between a set of response variables (y1, y2, y3...) and a set of predictor variables (x1, x2, x3,...)

Answer 56

Canonical correlation

Question 57

quadratic discriminat anaylsis + nd -'s

Answer 57

benefit is better group discrimination, the cost is that there is often 'over-fitting' so that while the discrimination works well for the 'training' data, it works less well in predicting group membership for new data.

Question 58

which stat method would you follow with null and saturated models to compare to?

Answer 58

logistic reg

Question 59

when would you want to change subject to a factor in R (subject

Answer 59

If this was a repeated-measures analysis, with subject as a random effect

Question 60

what is the order or columns and rows in R?

Answer 60

rows come first and columns second in R

Question 61

if the model fits well then this should be not much larger than two times the residual degrees of freedom ... which stat does this refer to?

Answer 61

logistic regression

Question 62

what does a poor fit in logisitic regression imply?

Answer 62

First, the response may not be related to x. Second, the relationship between the logit and x may be non-linear Third, there could be other variables affecting the response

Question 63

what is a null model? | what is saturated model?

Answer 63

x has no effect, so proportions same for all values of x model fits the data perfectly. The proportions are allowed to vary independently for every value of x. Fit x as factor not covariate

Question 64

when would you use chi sq over f test anova?

Answer 64

when comparing deviances rather than mean squares

Question 65

what is a loglinear model

Answer 65

loglinear model’ fitted in the same way as a logistic regression, just with predictors that are factors not continuous