PSYC206: Research Design and Statistics II

VARIABLE
○ Defined as: Anything that can change or have different values
○ Situational variables e.g., loud or quiet environments
○ Response variables e.g., reaction time, helping behaviour
○ Participant variables e.g., personality

○ Potential cause in the relationship

INDEPENDENT ○ Values independent of those of the other variables
VARIABLE (IV)
○ It can be manipulated or selected

○ “Consequence” in the relationship

DEPENDENT VARIABLE ○ DV is measured during/after exposure to the IV
(DV)
○ Its value depends on the value of the IV

BETWEEN SUBJECTS WITHIN SUBJECTS

Different people in each condition Same people in each condition

Manipulation of IV Manipulation of IV
EXPERIMENTAL
Random assignment to one condition Every participant go through each condition
DESIGNS

between subjects designs: (INDEPENDENT GROUPS)

ADVANTAGES DISADVANTAGES

Less powerful - Requires more participants

No order effects

No carryover effects Confounds due to individual differences across groups

○ Randomisation: Control for individual differences → Does NOT reduce variability

○ Minimising random variability → Similar participants; Large sample size
 PSYC206: Research Design and Statistics II

WITHIN subjects designs: (REPEATED MEASURES)

ADVANTAGES DISADVANTAGES

Reduced variability due to individual differences People take part in multiple conditions

More powerful - Requires fewer participants Order, practice, and boredom effects

○ Preventing these from becoming confounds

￫ Counterbalance: Generate possible orders and assign subjects randomly to each
↳ Order A: no games, games
↳ Order B: games, no games

Measurement scales
○ Determines the statistics we can use

SCALES OF MEASUREMENT

○ Label and categorise observations

○ No order/ mathematical relationship of categories
NOMINAL
○ A nominal measure used as a DV obtains frequencies
○ E.g., Sex (male, female, other), Levels of an IV

○ Categories organised in an ordered sequence

ORDINAL ○ Intervals are not equidistant
○ E.g., Grade (NN, PA, CR, DI, HD), Places in a race

○ Ordered categories
○ Equidistant intervals
INTERVAL
○ An arbitrary zero-point - no true ‘zero’
○ E.g., Temperature scales, Most psychological scales

○ Ordered categories
○ Equidistant intervals
RATIO
○ An absolute zero point - true ‘zero’
○ E.g., Height, Distance, Time
 PSYC206: Research Design and Statistics II

Types of statistics

1. Descriptive → Describe essential characteristics of data

2. Inferential → Infer characteristics of the population from the characteristics of a sample

Descriptive statistics

○ Frequency distribution
￫ Graph which plots the frequency of a score occurring
○ Normal distribution
DISTRIBUTION ￫ Values are evenly distributed around a central point
￫ Many statistical tests assume a normal distribution
￫ Symmetrical, bell-shaped and unimodal

○ Lack of symmetry
￫ Data points are not evenly distributed around a central point
○ Positive skew: Data clustered at the lower end
￫ Tail points toward the positive (right) end of the X-axis
○ Negative skew: Data clustered at the higher end
SKEWED DISTRIBUTION ￫ Tail points to the negative (left) end of the X-axis

○ The degree to which scores cluster at the ends (i.e., tails) of a distribution and
how pointy the distribution is
KURTOSIS DISTRIBUTION ○ Positive kurtosis (leptokurtic distribution)
￫ Long tails, pointy distribution
○ Negative kurtosis (platykurtic distribution)
￫ Short tails, flat distribution

○ Unimodal: One prominent high point

DISTRIBUTION MODAL
○ Bimodal: Two prominent high points
CHARACTERISTICS
○ Multimodal: Multiple prominent high points
 PSYC206: Research Design and Statistics II

○ Mean: Average score

CENTRAL TENDENCY ○ Median: Middle score
○ Mode: Most frequent score

○ Dispersion refers to the variability or spread of scores in the data

○ Measures of dispersion:
￫ Range: Highest value - Lowest value (e.g., 12 - 3)
￫ Interquartile range: IQR = Q3 - Q1
￫ Sum of squares: SS = Σ (X - μ)²
Σ (𝑋 − 𝑀)²
DISPERSION/ ￫ Variance: s² = 𝑛−1
VARIABILITY
Σ (𝑋 − 𝑀)²
￫ Standard deviation: s = 𝑛−1
 PSYC206: Research Design and Statistics II

Inferential Statistics

○ Frequency distributions determine the likelihood of a score occurring

PROBABILITY ○ The probability of a specific score is computed based on its frequency
DISTRIBUTION ○ Statisticians have calculated the likelihood of scores in a normal distribution with
a mean of 0 and a SD of 1
○ Any dataset can be normalised to a mean of 0 and a SD of 1 through z-scores

𝑋−µ
○ Z=
σ
○ Standardised score of a distribution
○ Indicates the distance between an X value and the mean
○ Forms a standardised distribution for comparisons between scores

○ 95% of the scores fall within

Z-SCORES
1.96 SD of the mean
○ Scores more than 1.96 SD
from the mean are rare
○ The same reasoning is
applied when computing test
statistics to assess the
likelihood of observing such a
large statistic if our null
hypothesis is correct

○ Inferences about the effects of a manipulation on the population, based on data

gathered from a sample
Steps for null hypothesis testing:
1. Generate hypotheses (H0 and H1)
￫ Null Hypothesis – H0: μ1 = μ2
NULL HYPOTHESIS
￫ Alternate Hypothesis:
TESTING
↳ Nondirectional – H1: μ1 ≠ μ2
↳ Directional – H1: μ1 > μ2
2. Select critical region → depends on significance level (α = .05)
3. Gather data and compute statistic (e.g., t test)
4. Make a decision (reject H0 or fail to reject H1)
￫ Obtained statistic = greater than the critical value: reject null hypothesis
￫ Obtained statistic = less than the critical value: retain null hypothesis
 PSYC206: Research Design and Statistics II

○ Assesses the evidence against a null

hypothesis
○ Indicates the probability of obtaining
P- VALUE = observed results due to random chance
PROBABILITY assuming the null hypothesis is true
￫ A smaller p-value suggests stronger
evidence against the null hypothesis
￫ Common significance thresholds are 0.05 or 0.01.

○ Two-tailed test = more ‘rigorous”

￫ Alpha (α) is split across both tails
￫ Allows for extreme values in either direction
ONE AND TWO-TAILED ￫ Harder to get a significant result
TESTS ￫ Smaller area beyond criterion (α)
○ One-tailed test = more “sensitive”
￫ Alpha (α) is only in one tail
￫ Easier to find a significant result
￫ Larger area beyond criterion (α)
○ Unless there is a strong directional prediction it is more appropriate to use
two-tailed tests

○ Type I Error: Rejecting H0 when it is true (false positive)

￫ The probability of this occurring is known as α-level (α = .05)

○ Type II Error: Accepting H0 when it is false (false negative)

￫ The probability of this occurring is known as β-level
￫ Depends on how much power you have to detect your effect

ERRORS IN ○ 1 - β is power
HYPOTHESIS TESTING ￫ Probability that a test will correctly reject a false null hypothesis (H0)
￫ Depends on sample size, effect size, standard error etc.
 PSYC206: Research Design and Statistics II

1. Independent Sample T-test

￫ Between-subjects design
￫ Compares two sample means from different groups
￫ Measure two groups on the same DV and compare means
T-TESTS 2. Paired Sample/ Repeated Measures T-test
￫ Within-subjects design
￫ Compares two sets of scores from the same participants
￫ Measure each participant twice on the same DV
￫ Examine the average difference in scores

○ T tests involve a categorical IV and a continuous DV.

○ Examine the link between two continuous variables.
○ Assess individuals on both variables to see if
being above/below the mean on one (e.g.,
CORRELATION
height) correlates with the other (e.g., weight).
○ Correlations reveal the direction and strength:
○ Direction is + or -.
a. Positive: As X increases, Y increases.
b. Negative: As X increases, Y decreases.
○ Strength is closer to |1| for a strong relationship, closer to 0 for a weak one.
 PSYC206: Research Design and Statistics II

INFERENTIAL STATISTICS

Probability Distribution:

○ Frequency distributions determine the likelihood of a score occurring.

○ The probability of a specific score is computed based on its frequency.
○ Statisticians have calculated the likelihood of scores in a normal distribution with a mean
of 0 and a SD of 1.
○ Any dataset can be normalised to a mean of 0 and a SD of 1 through z-scores

Z-SCORES

○ Z-Score = Standardised score of a distribution

○ Indicates the distance between an X value and the mean (in SD units)
○ Specifies the exact location of each X value within a distribution
○ Forms a standardised distribution for comparisons between scores

○ In a normal distribution, 95% of the scores are going to fall within 1.96 SD of the mean.
○ Scores more than 1.96 SD from the mean in either direction are rare.
○ We apply the same reasoning when computing test statistics to assess the likelihood of
observing such a large statistic if our null hypothesis is correct.
 PSYC206: Research Design and Statistics II

NULL HYPOTHESIS TESTING

○ Inferences about the effects of a manipulation on the population, based on data gathered
from a sample
Steps for null hypothesis testing:

1. Generate hypotheses (H0 and H1)

￫ Null Hypothesis – H0: μ1 = μ2

￫ Alternate Hypothesis: (Nondirectional – H1: μ1 ≠ μ2 | Directional – H1: μ1 > μ2 )

2. Select critical region → depends on significance level (α = .05)

3. Gather data and compute statistic (e.g., t test)

4. Make a decision (reject H0 or fail to reject H1)

￫ Obtained statistic = greater than the critical value: reject null hypothesis
￫ Obtained statistic = less than the critical value: fail to reject null hypothesis

P-VALUE = PROBABILITY

○ Assesses the evidence against a null hypothesis

○ Indicates the probability of obtaining observed results due to random
chance assuming the null hypothesis is true
￫ A smaller p-value suggests stronger evidence against the null hypothesis

ONE AND TWO-TAILED TESTS

○ Two-tailed test = more ‘rigorous”

￫ Alpha (α) is split across both tails
￫ Allows for extreme values in either direction
￫ Harder to get a significant result
￫ Smaller area beyond criterion (α)

○ One-tailed test = more “sensitive”

￫ Alpha (α) is allocated to one tail
￫ Easier to find a significant result
￫ Larger area beyond criterion (α)

Unless there is a strong directional prediction it is more appropriate to use two-tailed tests
 PSYC206: Research Design and Statistics II

Errors in hypothesis testing:

○ Type I error occurs when a true null hypothesis (H0) is incorrectly rejected (false positive)
￫ The probability of this occurring is known as α-level (α = .05)

○ Type II error occurs when a false H0 isn’t rejected (false negative)

￫ The probability of this occurring is known as β-level
￫ Depends on how much power you have to detect your effect

○ 1 - β is power
￫ Probability that a test will correctly reject a H0
￫ Depends on sample size, effect size, standard error etc.

T-TESTS

○ A t-test determines whether there is a statistically significant difference between the

means of two variables

1. Independent samples t-test:

￫ Compares two means from different participant groups (between-subjects design)
￫ Measure two groups on the same DV and compare means

2. Paired samples/ Repeated measures t-test:

￫ Compares two sets of scores from the same participants (within-subjects design)
￫ Measure each participant twice on the same DV
￫ Examine the average difference in scores
 PSYC206: Research Design and Statistics II

○ T-tests compare two groups and produces a t-statistic

○ Critical value from a t-distribution is used to determine statistical significance
○ P-value indicates likelihood of observed results with no real difference
○ If p-value is below 0.05, null hypothesis (no difference) is rejected
○ Rejection implies support for the alternative hypothesis (a significant difference)

Correlation:

○ With t tests, we have a categorical IV and a continuous DV

○ The relationship between two different continuous variables
○ Measure each participant on each variable
○ Examine whether being above/below the mean on one
variable (e.g., height) is associated with being above/below
the mean on the other variables (e.g., weight)
○ Correlations tell us the direction and strength of the
relationship between two variables
￫ Direction is given by sign of the correlation coefficient (+ or -)
↳ Positive (as X increases, Y increases)
↳ Negative (as X increases, Y decreases)
￫ Strength is given by the absolute value (magnitude) of the correlation coefficient
↳ Closer to |1|, stronger relationship
↳ Closer to 0, weaker relationship