Class Notes (1,100,000)

CA (620,000)

U of A (10,000)

PSYCO (1,000)

PSYCO104 (400)

Geoff Hollis (60)

Lecture 14

This

**preview**shows half of the first page. to view the full**2 pages of the document.**October 8 2014

TL-11

Random vs Stratified Sampling

Stratified Sampling is sometimes used in place of random sampling when there

are low-probability subgroups you are interested in explicitly representing.

I.e. Sampling age groups -- if there are 30% young adult Canadians,

systematically ensure that your data set includes 30% young adult Canadians.

Within the subgroup, return again to random sampling.

Making Claims

Correlation tests are used to detect whether two types of measures are

related

to each other.

When a change in one variable is accompanied by a proportional change

in

another variable, we say they are correlated.

E.g. Does aggression increase with frequency of video game

playing

Correlations can be visualized with scatterplots.

Statistical tools exist to determine strength of correlation. "The

correlation

coefficient" referred to as 'r' gives an estimate of strength.

0 means no correlation. Range of 1 to -1.

Causation - Observing correlations is not sufficient to establish cause and

effect. For instance, observing a correlation between spanking and adult

deliquency is not enough to establish that spanking your child causes them to

be maladapted.

There can be four types of reasons why you might observe a correlation.

X causes Y

Y causes X

X and Y are unrelated, but both depend on some third variable.

Or just random happenchance.

Experiments

Set up two identical situations, with only one change. If

differences are

observed, it must have been caused by the change which was

introduced.

In practice, it is difficult to only change one thing.

Random Assignment

###### You're Reading a Preview

Unlock to view full version