PSYC 333 Lecture Notes - Lecture 13: Observational Error, Total Variation, Dependent And Independent Variables
7 May 2018
School
Department
Course
Professor
PSYC 305 – STATISTICS FOR EXPER DESIGN, WINTER 2018
Lecture 13
this is hat e’e learnt in the past 2 eeks so far..
e’ll talk aout another speial ase of linear regression
focus is on comparing group means like in ANOVA.
subjects are random sample and randomly assigned to diff treatment groups.
this random assignment leaves individual backgrounds uncontrolled (e.g., u have 2 groups, but
because of random assignment, 1st group may have more females or younger people etc than
2nd group!)
no ontrol for gender/age differenes, et…
all due to random assignment
Analysis of Covariance
Linear Regression
We have thus far learned linear regression with
Continuous IVs
Nominal IVs with multiple levels (dummy-coded IVs)
We will learn a special type of linear regression, which has continuous
and nominal (dummy-coded) IVs at the same time.
Extraneous Variable
In experiments, subjects are randomly assigned to experimental
treatments to avoid any systematic bias in the treatment means.
However, random assignment leaves individual differences
uncontrolled.
Subjects may not be well matched on background factors.
Such extraneous variables or individual characteristics of subjects are
also likely to affect the dependent variable.
focus is on comparing group means like in ANOVA.
subjects are random sample and randomly assigned to diff treatment groups.
this random assignment leaves individual backgrounds uncontrolled (e.g., you have 2 groups, but
because of random assignment, 1st group may have more females or younger people than 2nd
group! o otrol for geder/age differees, et…all due to rado assignment.
find more resources at oneclass.com
find more resources at oneclass.com
PSYC 305 – STATISTICS FOR EXPER DESIGN, WINTER 2018
Example 1
here we have 2 groups, (gender is our key independent variable), we want to test effect of gender
on salary. we compare 2 group means. if there’s a diff, we may say yes there is an effect. can
apply t-test or one-way ANOVA. (comparing 2 parameters), if apply t-test there’s some
assumptions to be satisfied in advance like normality. By applying one of these tests, we can
compare these 2 group means.
but WORK EXPERIENCE is not controlled for. this is our EXTRANEOUS VARIABLE.
our main interest is to compare these 2 group means to see GENDER effect, but now these 2
groups have unequal means in work experience (females have lower work experience), so even
though there is difference in salary, we can’t be sure!!
work experience is not well matched because of random assignment & we know it affects salary.
we want to control for the effect of work experience.so use ANCOVA
find more resources at oneclass.com
find more resources at oneclass.com
PSYC 305 – STATISTICS FOR EXPER DESIGN, WINTER 2018
Example 2:
here we have 3 group means and want to compare these 3 group means on French test scores.
one-way anova (comparing more than 2 groups),
Analysis of Covariance
ANalysis of COVAriance (ANCOVA) is used to statistically control or
remove the influence of such extraneous variables on the dependent
variable.
More precise test of the difference among group means.
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
2nd group!) no (cid:272)ontrol for gender/age differen(cid:272)es, et(cid:272) all due to random assignment. We have thus far learned linear regression with. We will learn a special type of linear regression, which has continuous and nominal (dummy-coded) ivs at the same time. In experiments, subjects are randomly assigned to experimental treatments to avoid any systematic bias in the treatment means. Subjects may not be well matched on background factors. Example 2: here we have 3 group means and want to compare these 3 group means on french test scores. one-way anova (comparing more than 2 groups), Analysis of covariance (ancova) is used to statistically control or remove the influence of such extraneous variables on the dependent variable. More precise test of the difference among group means. The uncontrolled extraneous variables are called the covariate or concomitant variables in ancova. Ancova is an anova with covariate variables controlled. One is to control the influence of covariates on the.
