# PSY230H5 Lecture Notes - Lecture 2: Shyness, Convergent Validity, Discriminant Validity

14 views9 pages

Lecture 2

Sept,14,2015

Fundamentals II: Causality and Validity

Relevance of this Lecture

This lecture explains the relation between correlations and causality.

This lecture explains the distinction between personality measures and personality

constructs (validity).

Iclicker question:

Correlation doesn’t prove causality!

A = true, B = false

The Bermuda Triangle of Causality

Correlation does not prove causality

r = p1 + p2 + p3 *p4

too many unknowns!

Unless we have done an experiment we cannot claim causality

Correlation doesn’t prove causality but that doesn’t mean that we can’t use correlations to

see if something is causal

Often it is not feasible (ethically immoral) to conduct experiments on personality so we

have to use correlations to conduct research

Correlations can occur for a number of reasons but there is a limited number of reasons

that can be the causal variable

Using a curve arrow between the two variables just means they are related it doesn’t say

that they are directly related

There are three reasons why two variables related:

o1) A is a cause of B eg. variation of height causes weight

o2) B causes A (both #1 and #2 can occur at the same time)

o3) another variable can cause the two variables to relate to each other

R=p1+p2+p3*p4

Height and weight

R=.5

There is a third variable that is causing the two variables to relate to each other

find more resources at oneclass.com

find more resources at oneclass.com

Height and hair length

Hair length and weight have a -0.2 correlation (p1)

oMen have a shorter hair and have more weight and women have more hair and

have less weight

oGender is the third variable

oGender and hair length have a -0.6 correlation (p3)

There are more women in the class than males

This correlation may vary depending on the culture

There are more women in the class than males

Gender and weight have a 0.3 correlation (p2)

.3 x-.65=.195 which is p1(-.2) : shows that p2 x p3=p1

Reliability

Meaning of shyness varies from person to person

Shyness in the first week and shyness in the second week are a 0.7 correlation

Shyness and our memory of shyness last week has a 0.8 correlation

What we want u to forget ur response last week so that we can make sure that ur shyness

this week and last week is related by a third variable

When p3 and p4 are the same then there is only one unknown

The reliable variance explained 70% of variance in the first shyness test – means that we

have a stable sense of our shyness between the two weeks, we have a strong sense of self

When there is a reliability of 1 then p1 and p2 are directly relative

Reliability

Reliability coefficients assume that measurement errors are independent.

If u measure the same thing again it should be the same – reliability

Error at time 1 is unrelated to error at time 2.

Uncorrelated errors are called random errors.

If error variance is random, reliability coefficients estimate the amount of reliable

variance that is not due to random measurement error.

Whenever we do not find a perfect correlation we have to wonder why the data has

changed

Reliability =reliability x reliability^2

Two Methods to Estimate Reliability

find more resources at oneclass.com

find more resources at oneclass.com

Internal Consistency: Internal consistency can be measured by assigning half of the items

of a scale to one variable, and the other half to another variable, and then compute the

correlation between the two variables (split-half reliability).

oAsk the same question but worded differently

oAre you reserved vs. are you shy have a 0.5 correlation

Retest Correlation: Measuring the same items twice over a relatively brief period of time

(e.g., one week apart).

oProblem- we have to assume that things didn’t change between the two times

o

Problems

Internal Consistency:

- can be influenced by responses to previous items

- can be influenced by shared error variance

(e.g., something that happened that day)

Retest Correlations:

- can still be influenced by memory of previous response

- the true score could have changed over time

Important Implications

Measurement error lowers the ‘true’ correlations between two variables.

As a rule of thumb, observed effect sizes underestimate true effects.

For example, a correlation of .2 (r2 = 4%) could be a true correlation of .29 (r2 = 8%) if

the reliability of the two measures were .70.

Measurement error should be reduced whenever possible and your report must account

for the measurement error in your results

Validity

Is the measure that you are using the correct measure to test your construct

The relationship between a measure of height and the construct of height

Definition:

A measure measures the intended construct.

Examples:

A ruler is a valid measure of height.

A scale is a valid measure of weight.

The Relationship between a Measure of Height and the Construct of Height

find more resources at oneclass.com

find more resources at oneclass.com