1.3 Summary Slides on Review of Hypothesis Testing, Statistical Errors, and Effect Size

Question 1

Hypothesis testing

Answer 1

set of procedure used to assess the correctness of a hypothesis by examining sample data
the goal- to confirm or disconfirm predictions about differences among means (t, F), frequency/count data (chi square), or associations (r)

Question 2

two types of hypotheses

Answer 2

1. statistical Hypothesis
2. research Hypotheses

Question 3

Statistical Hypothesis

Answer 3

known as the null hypothesis -used to determines if the difference between means is reliably different where the null predicts that they are NOT different

Question 4

Research Hypothesis

Answer 4

the predicted difference among means that is based on your beliefs about the outcomes of your study.
Prediction based on theory
Ex: people who get the drug, compared to those in the control condition, will recall more items on a memory test

Question 5

The null and alternative hypothesis are:

Answer 5

The null hypothesis:
H0: µx = µy
The alternative hypothesis:
HA: µx ≠ µy

Question 6

Alpha Level

Answer 6

The alpha level (a) is the probability of committing a Type I error.
- Convention usually holds that the probability of occurrence of a type I error should be less than a 5% (means that the null hypothesis will be rejected less than 5% of the time if there is, in fact, no difference between means)

Question 7

Statistical significance

Answer 7

is determined by comparing the obtained statistical value (e.g., t statistics, F ratio) to a critical cutoff value
- accpeting or rejecting null hyp.
- If the obtained statistic (t or F) of greater that the critical cutoff value, then the null hypothesis is rejected, and differences are assumed to be true

Question 8

alt way to determine stat significance

Answer 8

If the obtained p value is less than our designated alpha level (p < .05) then the null is rejected, and significance is determined

Question 9

F-distribution

Answer 9

a one-tailed distribution
If the Fobt is greater than Fcrit then reject null
T2 = F

Question 10

Errors in Hypothesis testing

Answer 10

Type 1 error- eject the null hypothesis, when it is true
Mistakenly concluding that there is a difference between means when no difference exists
Type 2 error: when accept the null hypothesis, when it is false
mistakenly concluding that there is no difference between means when a difference exists

Question 11

Probability of committing Type 1 error is established by?

Answer 11

setting the alpha level (a)
- a of p< .05 means that the null hypothesis will be accidentally rejected 5% of the time
Thus, we have 95% confidence

-a of p<.01 means that the null hypothesis will be accidentally rejected 1% of the time
That gives us 99% confidence

Question 12

Probability of committing Type 2 error is established by?

Answer 12

beta (B)

Question 13

what is Power

Answer 13

The power of a test is a measure of one’s ability to reject the null hypothesis when it is false
Power is equal to 1 – beta
Power as an estimate ranges from 0 to 1.0, where 0 is no power and 1.0 is perfect power
Cohen recommends that power should be ≥ .8

Question 14

Factors that affect Power

Answer 14

-effect size
-sample size
-variablity in the measure
-choice of alpha level

Question 15

what does effect size describe?

Answer 15

the magnitude of the effect that an independent variable (factor) has on the dependent variable
-usually thought about in terms of small, medium, or large

Question 16

measuremnets of effect size

Answer 16

1. cohen’s d
   2.Eta squared
2. Omega squared

Question 17

Cohen’s D effect size

Answer 17

-is a measure of the magnitude of the differences between two groups measured in standard deviation units
-Cohen’s d is computed as the difference between two means divided by the standard deviation of al scores (pooled standard deviation)

Question 18

Eta Squared effect size

Answer 18

the proportion of variance in the DV scores explained by the IV
It is a biased estimate of effect size limited to the sample, it does not allow for generalization to the larger population
It is computed as the sum of squares treatment divided by the sum of squares total,
Where: .01 is a small effect; .06 is a medium effect; .14 is a large effect

Question 19

Omega Squared effect size

Answer 19

proportion of explained variance in the DV scores by the IV after controlling for other predictors
It is a less biased estimate of effect size that is not limited to the sample; it allows for generalization to the larger population
Where: .01 is a small effect; .06 is a medium effect; .14 is a large effect