PSYCHOLOGICAL STATISTICS FINALS Reference Data = Hypothesized Mean
T-TEST ➢ Test Value - hypothesized value of the mean in the
population. It comes from a literature review, trusted
ANOVA research organization, legal requirements, or industry
standards.
CORRELATION FOUR STEPS OF HYPOTHESIS TESTING
1. State the null and alternative hypotheses
REGRESSION ANALYSIS
2. Set the criteria for a decision (0.05 critical value)
CHI-SQUARE
3. Calculate the t test statistic using Jamovi
4. Make a decision (Accept or Reject)
General Terms
COHEN’S D
- Accompanies the hypothesis test to measure effect
size.
- The larger the size, the more powerful the study
T-TEST STATISTIC
Cohen’s D Strength Interpretation
- An alternative to z-score.
0.8 Large Effect
- Might be considered an “approximate” z.
- Used to test hypotheses when the value of σ is unknown.
0.5 Medium Effect
- Estimated standard error (sM) is used in place of the real
standard error when the value of σ is unknown. 0.2 Low Effect
SIGNIFICANCE LEVEL
Threshold probability to reject the null hypothesis, commonly PERCENTAGE OF VARIANCE
set at 0.05. - Determine the amount of variability in scores
- Alternative method for measuring effect size
DEGREES OF FREEDOM
- Number of scores in a sample that is “free to vary.” r2 = 0.01 Small effect
- Only n-1 scores in a sample are independent
- Noted as df r2 = 0.09 Medium effect
- df = n - 1
r2 = 0.25 Large effect
INDEPENDENT SAMPLE
PAIRED SAMPLES T-TEST
- Determine if there is statistically significant difference
between two independent groups.
- “BEFORE” and “AFTER”
Null Hypothesis assumes no difference in the group means - A.k.a Repeated-measures design/ within-subjects design
- Two separate scores are obtained for each individual in
Alternative Hypothesis assumes a difference exists the sample
- 1 Population with treatment
Pooled Variance - a weighted average of the variances of the - Same subjects are used in both treatment
two groups, used when variances are assumed equal. - Based on difference scores (D) rather than raw scores
(X)
Assumptions:
ONE SAMPLE T-TEST
1. Observations within each treatment condition must be
independent
Assumptions: 2. Population of difference scores must be normally
- Data values are; Independent, Continuous, distributed
- Obtained via a simple random sample from the data ➢ With relatively large samples (n > 30) this assumption can
- Population is assumed to be normally distributed be ignored
- Has reference data
ANALYSIS OF VARIANCE 3. You should have independence of observations, which
means there is no relationship between the
observations in each group or between groups
- Statistical formula used to compare variances across
themselves.
means or average of different groups.
4. There should be no significant outliers
- Determines if there is any difference between means of
5. DV should approximately be distributed for each
different groups.
category of each of the IV, alternatively, the residuals of
Statistical Hypothesis for ANOVA
the dependent is approximately normally distributed.
● Null Hypothesis: The level or value on the factor does
6. There needs to be homogeneity of variances.
not affect the dependent variable.
Alternate Hypothesis for ANOVA EFFECT SIZE FOR ANOVA
● H1: There is at least one mean difference among - Compute percentage of variance accounted for by the
populations treatment conditions
- Eta-squared = η²
F ratio = Determines if there is significant difference
- Based on variance instead of sample mean
differences η² = 0.01 Small effect
- Denominator of the F-ratio is called the error term
η² = 0.06 Medium effect
Variance (differences) between sample means
F = ------------------------------------------------------------------- η² = 0.14 Large effect
Variance (differences) expected with no treatment
ONE WAY ANOVA
P value = Decides if the null hypothesis is rejected or
accepted - 1 Independent with different levels, 1 dependent
Logic of ANOVA Null Hypothesis all group means are equal
● Between-treatments variance/ between group variability:
- Variability results from general differences between Alternative Hypothesis at least one group mean is different
conditions
- Variance between treatments measure differences F-Ratio - the ratio of mean square between groups to mean
among sample means square within groups
● Within-treatments variance/ within group variability: TWO WAY ANOVA
- Variability within each sample
- Individual scores are not the same with each sample
- 2 Independent with different levels, 1 dependent
SOURCE OF VARIABILITY BETWEEN TREATMENTS
- Systematic differences caused by treatments POST HOC TESTS
- Random, unsystematic differences - Additional tests done to determine exactly which mean
● Individual differences differences are significant, and which are not
● Experimental errors - Conducted after finding a significant F-value
- No systematic differences related to treatment groups
occur within each group Games-Howell Nonparametric statistical analysis that
post hoc test compares all possible differences
Analysis of degrees of freedom when the assumption of homogeneity
- Total DOF: dftotal = N -1 of variances is violated
- Within-treatments DOF: dfwithin = N - k
- Between-treatments DOF: dfbetween = k- 1 Tukey Assess the significance of differences
between pairs of group means
Assumptions:
1. Dependent Variables should be measured at the Equal Variances
interval or ratio (continuous).
2. Independent Variables should consist of two or more
categorical, independent groups.
CORRELATION 2. The relationship between variables is linear
3. Homoscedasticity (equal variance of residuals across
Correlation Coefficients values).
- The relationship between two variables. How the value 4. There are no significant outliers, as they skew the
of one variable changes correlation values.
- A correlation coefficient is a numerical index to reflect
NON PARAMETRIC
the relationship between two variables.
- Nominal / Ordinal
Coefficient Determination
★ SPEARMAN’S RANK CORRELATION: Used to
Represents the proportion of variance in 1 variable explained
measure the degree of association between two
by the other.
variables. Does not carry any assumptions about the
Range: -1~ + 1 distribution of the data and is the appropriate correlation
analysis when the variables are measured on a scale
The magnitude of the relationship is indicated by the that is at least ordinal.
absolute value of the coefficient, or the size of the number
Assumptions:
without regard to the sign.
1. Data must be ordinal
2. The scores on one variable must be monotonically
A higher absolute value indicates a stronger relationship.
related to one other variable
0: There is no linear relationship
+1 or -1: There is a perfect linear relationship or complete ★ KENDALL TAU RANK CORRELATION: Non-parametric
correlation between two variables correlation coefficient that measures strengths and
direction of association between two variables based on
Positive: Direct Relation Negative: Inverse Relation
the ranks. It evaluates the ordinal association by
comparing the number of concordant and discordant
↑ Variable 1 ↑ Variable 2 ↓ Variable 1 ↑ Variable 2 pairs in the data.
Assumptions:
EX. As the recitation(V1) EX: As the absence (V1) 1. Data should be ordinal or continuous.
increases), grades (V2) decreases, grades (V2) 2. No assumption of normality or linearity is required.
increases. increases. 3. Best suited for data that can be ranked and where
relationships are monotonic
PLOT: PLOT:
SCATTERPLOT
A graphical representation of the relationship between 2 Types of Linear Correlation
variables. ● Linear Relationships
➢ Strong relationship - close data points
➢ Weak relationship - scattered data points
PARAMETRIC
● Curvilinear Relationships
- Pearson product-moment correlation (Name after
inventor Karl Pearson Interval Coefficient Relationship Level
- Interval / Ratio
- Continuous Data (Height, age, test score, income) 0.80 — 1.000 Very strong
★ PEARSON R CORRELATION: most widely used 0.60 — 0.799 Strong
correlation statistic to measure degree of relationship
between linearly variables 0.40 — 0.599 Moderate
0.20 — 0.399 Weak
Assumptions of Pearson r:
1. Both variables are continuous and normally distributed.
0.00 — 0.199 Very Weak
p-Value
REGRESSION ANALYSIS Indicates the statistical significance of the model or individual
predictors. Smaller values (<.05) suggest significance.
- Statistical method used to understand relationships
between variables Residuals
The difference between observed and predicted values
Objectives:
- Predict outcomes (DV) using one or more predictors (IV) Standard Error of Estimate (SEE)
- Understand the strength and direction of relationships Measures the average size of the residuals; smaller values
between variables indicate better fit.
Assumptions:
TYPES OF REGRESSION ANALYSIS 1. Linearity: the relationship between the IV and DV is
linear
2. Homoscedasticity: the variance of the residuals is
Simple Linear Regression: One independent variable constant across all levels of X
3. Independence: observations are independent of each
Multiple Linear Regression: 2 or more independent other
variable 4. Normality: residuals (error) are normally distributed
5. No Multicollinearity: predictors are not highly
correlated with one another
Logistic Regression: For binary or categorical outcomes
Hierarchical and Moderated Regression: Advanced
F-Value Associated Interpretation Remarks
techniques in psychology p-Value
DEPENDENT VARIABLE (Y) Close to 1 p>.05 No significant Model
the outcome or response variable being predicted or relationship explains no
between more
explained
predictors and variances
outcome than would be
INDEPENDENT VARIABLE (X) expected by
The predictors or explanatory variable/s used to predict the chance.
dependent variable
Between 2-10 p<.05 Moderate Predictors
INTERCEPT (B) evidence that explain some
The predicted value of Y when all X values are zero; the the model is variance but
baseline model significant the effect size
may be small.
SLOPE
Greater than 10 p<.05 Strong Predictors
Represents the rate of change in YYY for a one-unit change
evidence that explain a
in XXX the model is substantial
significant proportion of
COEFFICIENT (B1,B2,...) the variance
The slope of the of the regression line, representing the in the DV.
change in Y for a one-unit increase in X
Very high (F>50) p<<.05 Very strong Typically seen
2
R (Coefficient of Determination) evidence that in well fitting
The portion of variance in the dependent variable explained the model is models with
by the model. significant strong
predictors or
large sample
Adjusted R2 sizes.
Adjusted for the number of predictors to avoid overestimating
R2 in models with many variables.
F-Statistic
Tests whether the overall regression model is statistically
significant
CHI-SQUARE TEST Magnitude of Larger Residuals indicate a greater
the Residual contribution to the Chi-Square statistic
- A non-parametric statistical test
For standardized residuals:
- Compare observed frequencies with expected
Residuals > I 2 I are typically considered
frequencies.
significant contributors.
- Determines whether there is a significant association
between variables.
- Commonly used tests: Chi-square Test of Independence
and Goodness-of-Fit. THE GOODNESS-OF-FIT TEST
IMPORTANCE OF CHI-SQUARE - Determines whether the observed frequencies of a
- Evaluate relationships between categorical variables single categorical variable match expected frequencies
- Widely used in psychological studies to test hypotheses under a specific hypothesis.
- Helps in identifying patterns or associations in data - Evaluates how well the observed distribution of data fits
- Provides insight into preferences, behavior trends, or an expected theoretical distribution.
demographic differences.
WHEN TO USE:
1. When you have categorical variable
Assumptions:
2. To compare observed frequencies with expected
1. Data must be in frequencies (not percentages or
frequencies
continuous values)
2. Observations must be independent Examples:
3. Expected frequencies should be at least 5 for all ● Check if observed patient preferences align with
categories expectations
4. Variables must be categorical ● Assess whether response patterns in an experiment
match theoretical predictions
5. No overlapping counts in data
6. The sample should be randomly selected.
Observed Frequency (OOO) Residual in Goodness-of-Fit Interpretation
The actual count in each category Large residuals indicate that the observed frequency differs
significantly from the expected frequency
Expected Frequency (EEE)
The theoretical count based on the null hypothesis ● Positive residual: observed frequency exceeds the
expected frequency
Residuals
- Residuals in a Chi-Square test represent the ● Negative residual: observed frequency is less than
differences between observed frequencies and expected
expected frequencies in a contingency tables
- Helps identify patterns or deviations in the data that Criteria for High/Low Residuals
contribute significantly to the overall Chi-Square > 2 or < -2 Generally considered significant at the
statistic. α=0.05 level
IMPORTANCE OF RESIDUALS > 3 or < -3 Highly significant
- Detectives like– they help figure out where data is
behaving in an unexpected way.
- If the residuals are large, it means there’s a bigger CHI-SQUARE TEST OF INDEPENDENCE
difference between what we observed and what we
expected
- Determines whether a population’s two categorical
Interpreting Residuals variables are associated or independent
- Evaluates whether the distribution of one variable
Sign of the Positive Residual: Observed frequency depends on another.
Residual is higher than expected
● O>EO > EO>E WHEN TO USE :
1. When you have two categorical variable
Negative Residual: Observed 2. To analyze whether variables are related
frequency is lower than expected
● O<EO < EO<E
Type I Error
Examples:
● Determine if therapy preferences vary by age, gender, or Incorrectly rejecting the null hypothesis when it is true
other demographics
● Analyze relationships between diagnosis and treatment Type II Error
adherence Failing to reject the hypothesis when it is false
Power
Residual in Independence Interpretation The probability of correctly rejecting a false hypothesis
Residuals highlight which cell(s) contribute most to the
Chi-Square stats and hence to the rejection of independence
● Positive residual: indicates an
overrepresentation of the observed frequency
compared to the expected
● Negative residual: suggests underrepresentation
Indicates significant
Adjusted residuals > 2 or <-2 deviation from expected
independence
Interpreting Results
● High Chi-Square Statistic values suggest stronger
association
● If p <0.05, reject null hypothesis
● Analyze residuals to identify specific categories
contributing to the results
● Report finding clearly, including degrees of freedom (df)
and sample size.
Applications in Psychology
● Understanding relationships between demographic
factors and behaviors
● Analyzing survey data (ex. Preferences, attitudes)
Good luck po so much…..
● Studying clinical diagnoses and treatment categories
● Examining group differences in experimental studies
Common Pitfalls to Avoid
● Ignoring assumptions
● Misinterpreting causation from associations
● Using Chi-square for continuous data
● Not checking for independence of observations
General Statistical Terms
P-Value
The probability of observing a test statistic as extreme as the
one calculated, under the assumption that the null hypothesis
is true.
Effect Size
A measure of the strength of a phenomenon, independent
sample size.