Intro to Hypothesis Testing 3.3 & 4.1

Question 1

What do statistical tests evaluate?

Answer 1

Statistical tests evaluate the likelihood that the differences between the
sample means reflect a real difference in the population, i.e., that the null
hypothesis should be rejected

Question 2

What is Alpha?

Answer 2

sets criterion for decision, set by researcher, indexes critical value (typically 0.05 in cogsci) (for two tailed test alpha/2 for each side)

Question 3

Different sampling methods:

Answer 3

Random sampling: rare
Convenience sampling: common, external validity concerns, WEIRD(Western, educated,
industrialized, rich, democratic )

Question 4

Parameter

Answer 4

a characteristic of a population

Question 5

Parametric tests

Answer 5

•Ratio & interval data
•Meets assumptions about the distribution of the parameter
-the DV is normally distributed
-variance is homogenous
-Independence of observations

•Generally, more powerful than nonparametric tests

Ex: T-test (only two groups being compared), ANOVA: more than two groups

Question 6

Non-parametric tests

Answer 6

If the DV is nominal or ordinal, or if the data is
not normally distributed. No assumption about the distribution of the parameter.

Ex: chi square test (nominal data), Mann Whitney U test (ordinal data)

Question 7

Error

Answer 7

e element of variability produced by extraneous factors, such as
measurement imprecision, that is not attributable to the IV or other controlled
variables

Question 8

Comparing variability

Answer 8

Comparing the variability within and between groups in your sample is the
basis for making inferences about the population

(dif between groups/ dif withing groups) = ((effect of IV +error)/ error)) = test statistic

Question 9

Non-significant difference

Answer 9

• The numerical difference between the means is likely due to chance
• Fail to reject the null hypothesis.
• Fail to support the alternative hypothesis
• Note: You DO NOT accept the null hypothesis if you fail to reject it

Question 10

Significant difference

Answer 10

• The numerical difference is unlikely due to chance
• Reject the null hypothesis
• Find support for the alternative hypothesis
• Note: You DO NOT prove the alternative hypothesis if you reject the null

Question 11

p value

Answer 11

The probability of obtaining results if the null hypothesis was true. Reject the null hypothesis if its lower than alpha

Question 12

p value and alpha relation

Answer 12

p<a>a nonsignificant</a>

p=a or slightly above a, marginally significant</a>

Question 13

Type 1 error

Answer 13

False positive
ex: rejecting null hypothesis ( in theory its true)
Alpha error

Question 14

Type 2 error

Answer 14

False negative
ex:fail to reject null hypothesis ( in reality it’s false)
Beta error

Question 15

Beta

Answer 15

1-b= Power (the likelihood of finding a true difference if one exists)
- Underpowered studies are relatively unlikely to detect a difference even if one exists

Question 16

How to increase statistical power?

Answer 16

• parametric more powerful than non-parametric
• take more measurements
• include more participants
• more observations