You are on page 1of 24

ANALYSIS OF

(CONTINUE
VARIANCE
D) Lecture 5
February 4,
2019
One-way
ANOVA
■ Simplest version of ANOVA
– Testing for difference in means between more
than two groups
– Generally have one independent variable
(group or
factor) and one dependent variable.
Repeated measures
ANOVA
■ Used when there are multiple measurements of
the same variable from the same study
subjects
– E.g. ?
■ Same principles as one-way ANOVA, but
considers dependency between multiple
measurements
Multivariate
ANOVA
(MANOVA)
■ Evaluating effects between an independent
(predictor) variable and two or more dependent
variables (outcomes)
– Predictor/Independent variable: biological
sex
– Outcomes/Dependent variables: body mass
index and waist circumference
– Can have several dependent variables (more
than 2) if sample size permits
MANOVA
output
vs. ANOVA (one way)
output
MANO
VA
■ Can protect against Type I error
■ In some cases, combination of dependent
variables better represents a phenomenon than
one single dependent variable
■ May reveal differences not discovered by ANOVA
■ Drawbacks:
– More complicated design
– Lose DF for each dependent variable
included
Factorial
ANOVA
■ Interested in the effects of more than one factor
(independent variable) together on an outcome
– Factor 1 (Diet): Low carb (<25% energy) vs.
very low carb diet (<10% energy)
– Factor 2 (Gender): Male vs. female
– Outcome: Weight loss
■ Yields a 2x2 factorial design -> requires two-way
ANOVA
– Group A1: Low carb diet on male subjects
– Group A2: Very low carb diet on male subjects
– Group B1: Low carb diet on female subjects
– Group B2: Very low carb diet on female subjects
■ Enables assessment of interactions between
factors on outcome
Interaction
effects
■ Simultaneous effect of two (or more) independent
variables on a dependent variable. The joint
effect is significantly greater (or significantly less)
than the sum of the parts (i.e. than the individual
main effects).
■ Thinking of the two-way ANOVA example:
– Diet: main effect of diet on weight loss
– Gender: main effect of gender on weight loss
– Diet*Gender: Interaction effect of diet and
gender
together on weight loss
http://statisticsbyjim.com/regression/interaction-
effects/
Example with CCHS
database

■ What interaction effect are we evaluating with


this model?
Analysis of Covariance
(ANCOVA)
■ Used when you want to control for effects of
other relevant variables
– Adjust for covariates
– Helps to mitigate potential for confounding

■ Will discuss ANCOVA in regression lecture


Additional
considerations
about
■ ANOVA
Fixed vs. random factors
– Fixed: Assumption that values of a fixed
variable in one study are the same as the
values of the fixed variable in another study.
– Random: Assumption that values are drawn
randomly from a larger population of value,
and therefore, represent that population.
■ Factors are assumed to be fixed in most ANOVA
analyses
■ Contrasts
– Test for a particular pattern of differences
among the means
– More than two groups can be compared at
once
Harding et al. 2010
Article
■ What was the purpose of this
article?
■ What was new about this
article?
■ Briefly, what were the
methods?
– How many animals?
– How many study groups?
■ What types of statistics did
they use?
■ What were the conclusions?
Assignment
notes
■ Find your partner for the project
■ Review poster tips and peer feedback form
■ Get to know your variables of
interest in your NHANES dataset
■ Refer to this website to obtain information
on your variables

https://wwwn.cdc.gov/nchs/nhanes/con
tinuo usnhanes/default.aspx
Steps when inspecting
variables
■ Examine the distribution of your variables
and note your sample sizes
■ Think of any variables you might
want to manipulate prior to
conducting your analysis
■ Do you need to transform any
variables?
■ Do you need to exclude certain
participants?
Midter
m2 hours in CC Ballroom (Feb 11, 11:35-1:35pm)

■ Total 40 marks: Multiple choice (/20) and short
answer (/20)
– Covers material from first 5 weeks of class
– No aids permitted
■ Please use a pencil for scantron, can also use
pencil for written answers.
■ Point form answers acceptable, but writing
must be legible.
Reading for Feb
20
■ Baddour et al.
2013

You might also like