DISCUSSION OF SIR EGONZ

DESCRIPTIVE STATISTICS - Focuses on summarizing and organizing data. It describes

the main features of a dataset through numerical summaries, graphs, and tables. Examples
include measures of central tendency (mean, median, mode) and measures of variability
(range, variance, standard deviation). Used when you just want to present or describe data
without making conclusions beyond the dataset.
INFERENTIAL STATISTICS - Goes beyond description and makes predictions or
generalizations about a population based on a sample. Uses probability theory to infer
conclusions and make decisions. Includes hypothesis testing (e.g., t-tests, chi-square tests,
ANOVA) and confidence intervals. Essential for making educated guesses about larger
populations based on limited sample data.
PARAMETRIC - Assumes that the data follows a known distribution (usually normal
distribution). Requires the data to meet certain conditions, such as homogeneity of
variance and independence. More powerful and precise when assumptions are met.
Examples of parametric tests:
 T-test (compares means of two groups)
 ANOVA (compares means of multiple groups)
 Pearson’s correlation (measures strength of a linear relationship)
  Linear regression (predicts a dependent variable from an independent variable)
Example: If you want to compare the average test scores of two different classes and you
assume the scores are normally distributed, you would use a t-test (parametric).
NON-PARAMETRIC - Does not assume a specific distribution of the data (good for skewed
or small datasets). More flexible and can be used with ordinal data or non-normal
distributions. Often used when assumptions for parametric tests are violated.
Examples of non-parametric tests:
 Mann-Whitney U test (alternative to t-test, compares two independent groups)
 Kruskal-Wallis test (alternative to ANOVA, compares more than two groups)
 Spearman’s rank correlation (alternative to Pearson’s correlation)
 Chi-square test (used for categorical data analysis)
Example: If you want to compare the median test scores of two classes, but the scores are
heavily skewed (not normally distributed), you would use the Mann-Whitney U test (non-
parametric).

Use parametric if your data follows a normal distribution and meets the assumptions.
Use non-parametric if your data is skewed, small, or doesn't meet assumptions.
BROAD CATEGORIES OF DATA
NOMINAL (Non-parametric) - Data consists of categories or labels without any meaningful
order. No numerical value or ranking. Used for classification.
Examples:
 Gender (Male, Female, Non-binary)
 Blood Type (A, B, AB, O)
ORDINAL (non-parametric) - Data is categorized with a meaningful order, but the
differences between categories are not uniform or measurable. Ranking exists, but the
gaps between ranks are not equal.
Examples:
  Survey Responses (Strongly Disagree → Disagree → Neutral → Agree → Strongly
Agree)
 Socioeconomic Status (Low, Middle, High)
INTERVAL (Parametric) - (Numerical - No True Zero) Data has a meaningful order with
equal intervals between values. No true zero (zero does not mean "none" or "absence").
Addition & subtraction make sense, but multiplication & division do not.
Examples:
 Temperature in Celsius or Fahrenheit (0°C does not mean "no temperature")
 IQ Scores (An IQ of 140 is higher than 120, but 140 is not "twice as intelligent" as 70)
RATIO (Parametric) - Data has a meaningful order with equal intervals and has a true zero
(zero means "none" or "absence"). Supports all arithmetic operations (addition, subtraction,
multiplication, division). Most precise level of measurement.
Examples:
 Age (A 40-year-old is twice as old as a 20-year-old)
 Height & Weight (A person weighing 80kg is twice as heavy as a person weighing 40kg)

DICHOTOMOUS - A dichotomous variable has only two possible categories or values.

Often used in yes/no, true/false, or pass/fail situations. Common in categorical data analysis
and logistic regression.
NON-DICHOTOMOUS - A non-dichotomous variable has three or more possible
categories or values. Can be nominal, ordinal, interval, or ratio data. More complex and
requires different statistical approaches.
Within the subjects (Dependent or Repeated Measures) - Same participants experience all
conditions of the experiment. Each person serves as their own control, reducing variability.
Requires fewer participants because comparisons are made within the same group. More
statistical power since individual differences are minimized. Risk of carryover effects (e.g.,
practice effects, fatigue, learning effects).
Examples:
 A memory experiment where participants first study words under bright light and then
study words under dim light (same people, different conditions).
  A drug study where each patient takes a placebo first, then the actual drug (same people
tested twice).
 A reaction time test where subjects use their dominant hand first, then their non-
dominant hand (same people, two tasks).
Think of it as "before and after" testing using the same group!
Between the subjects (Independent) - Different participants experience different
conditions of the experiment. No risk of carryover effects because each person is tested only
once. More variability due to individual differences. Requires a larger sample size to
balance individual differences.
Examples:
 A group of students is split into two, where one group learns using PowerPoint, and
another group learns using a printed book (different people in each condition).
 A drug trial where Group A receives the drug and Group B receives a placebo (two
separate groups, no crossover).
 A company testing two marketing strategies where one set of customers sees Ad A and
another set sees Ad B (different customers, different ads).
📌 Think of it as splitting people into separate groups, each experiencing only one condition!

MEASURES OF CENTRAL TENDECY

MEASURES OF CENTRAL TENDENCY - Measures of central tendency describe a dataset's
"typical" value they tell us where the center of the data lies.
OTHER DEFINITIONS – It refers to statistical values that represent the center or typical value
of a dataset. These measures help summarize large amounts of data by identifying a single
value that best describes the "average" or "middle" of the dataset.
MEAN – It is also known as average, where it is the sum of the scores divided by the
number of scores. It is considered parametric, and follows normal distribution assumption,
represented by the (µ). It is very sensitive to outliers or extreme values, and it is usually
used in variance in measuring the dispersion of the values in a data set and the difference of
values to the mean.
MEDIAN – It is the middle number in a dataset. The median is sometimes used instead of the
mean, especially if the data are skewed or has a presence of outliers. Non-parametric and
Ordinal, and immune to the effect of extreme values since we are just determining the
middle score.
MODE - It is the value that appears most frequently in the dataset. There could be more than
one mode in a dataset. It is considered non-parametric, categorical, and nominal. It is based
on frequency or percentage.

In summary: The mean is the most commonly used measure, but the median is better when
dealing with skewed data or outliers, and the mode is useful for categorical data.

MEASURES OF VARIABILITY (DISPERSION)

MEASURES VARIABILITY - In statistics, measure of variability or measures of dispersion,
describe how scores in a given dataset differ from one another. It also represents how well
the mean or the median represents the data, and how much we can trust the conclusion based
on mean and median. It is also one of the descriptive statistics that helps us summarize our
data.
OTHER DEFINITIONS - Measures of variability describe how spread out or dispersed the
data is around the central tendency (mean, median, or mode). They help in understanding how
much the data fluctuates or differs from one another.
RANGE - The range is the simplest measure of variability (or dispersion). States the
difference between the largest value and smallest value. It is highly affected by outliers
and does not consider how values are distributed.
INTERQUARTILE RANGE – It is a measure of dispersion, which is the spread of data. It is
defined as the difference between 25th percentile to the 75th percentile. The IQR is used to
measure the variability of the test if the data are skewed, and it is used median instead of
meaning in showing the measure of central tendency. It is a good alternative to variance, if
extreme outliers are present in the data since it uses media as a reference in determining the
central tendency.
Why is it a good alternative to variance?
 Variance is sensitive to extreme values or outliers because it squares the deviations
from the mean, which means that in skewed distributions, a few extreme values can
drastically inflate the variance.
 IQR is a better measure of variability for skewed data or data with outliers, as it’s not
influenced by extreme values like variance or standard deviation.
 Variance is more useful when the data is normally distributed and when you're okay
with the influence of outliers.
VARIANCE - Variance is the average squared difference of the values from the mean.
Measures how far each value in the dataset is from the mean. Best used when you need a
precise measure of spread, especially in normally distributed data.
STANDARD DEVIATION - Standard deviation is the standard or typical difference between
each data point and the mean. When the values in a dataset are grouped closer together, you
have a smaller standard deviation, and when the values are spread out more, the standard
deviation is larger because the standard distance is greater.

Summary:
 Range = Simple but affected by outliers.
 IQR = Best for skewed data because it ignores outliers.
 Variance = Good for precise calculations but harder to interpret.
 Standard Deviation = Best for normal distributions, easy to interpret.

MEASURES OF SYMMETRY
MEASURES OF SYMMETRY – Measures of symmetry describe whether a dataset is
symmetrically distributed or skewed. Symmetry means that the left and right sides of a
distribution (around the center) look the same. If the data is not symmetric, it is called
skewed.
SKEWNESS - Measures the asymmetry of data distribution. A skewness of 0 means the data
is perfectly symmetrical (normal distribution). Positive or negative values indicate
skewness.
Interpretation:
 Sk = 0 → Symmetric Distribution
 Sk > 0 → Positively Skewed (Right Skewed)
o Tail is longer on the right

o Mean > Median

 Sk < 0 → Negatively Skewed (Left Skewed)

o Tail is longer on the left

o Mean < Median

KURTOSIS - Measures the "tailedness" of the distribution (how extreme the outliers are).
Interpretation:
 K = 3 → Mesokurtic (Normal Distribution)
o Moderate tails, same as a normal distribution

 K > 3 → Leptokurtic (Heavy-Tailed Distribution)

o More extreme values (outliers), sharper peak

 K < 3 → Platykurtic (Light-Tailed Distribution)

o Fewer extreme values, flatter peak
SYMMETRICAL DISTRIBUTIONS - In a symmetrical distribution, the right side mirrors the
left. The median is at the center, dividing the area equally. The mean also lies at the center,
balancing scores on both sides.

SKEWED DISTRIBUTION - In skewed distributions, the mean, median, and mode are
usually in different positions, and their order can be predicted based on the direction of the
skew.
Why is Symmetry Important?
 Helps choose the right measure of central tendency:
o If data is symmetric, use mean.

o If data is skewed, use median.

 Affects statistical tests: Many tests (e.g., t-tests, ANOVA) assume normality, so
skewness and kurtosis help determine if normality assumptions are met.
Summary:
 Skewness measures how much a distribution is shifted left or right.
 Kurtosis measures how peaked or flat the distribution is.
 Normal distributions have Sk = 0 and K = 3.
 Skewed distributions impact data interpretation and statistical analysis.

T-TEST FOR INDEPENDENT SAMPLES

T- TEST - A t-test is a statistical test procedure that compares two mean values. It is often
used to determine if the valued means have any significant differences.
TYPES OF T-TESTS
ONE SAMPLE TEST - The sample mean is compared to a standard value. (Standardized
mean)
INDEPENDENT SAMPLE (BETWEEN GROUPS) - Two valued means are compared and
tested.
DEPENDENT (PAIRED SAMPLE OR WITHIN GROUPS) - Mean values are measured on one
sample group.
T-TEST FOR INDEPENDENT SAMPLES - The t-test for independent samples (also called
an independent t-test or unpaired t-test) is a statistical test used to compare the means of
two separate (independent) groups to determine if there is a significant difference between
them.
When to Use an Independent t-Test?
Use an independent t-test when:
✅ You have two different groups (e.g., male vs. female, experimental group vs. control group).
✅ You want to compare their means to check for a significant difference.
✅ The data is normally distributed and has equal variances (though Welch’s t-test can be
used if variances are unequal). (use Levene’s Test to check if the ).
✅ The groups are independent (i.e., no relationship between them, such as different people in
each group).
STEPS TO CONDUCT AN INDEPENDENT T-TEST
STATE THE HYPOTHESES

 Null Hypothesis (H₀): There is no significant difference between the means of the two
groups.

 Alternative Hypothesis (H₁): There is a significant difference between the means of the
two groups.
CHECK ASSUMPTIONS
 The data should be independent (from different participants).
 The data should be normally distributed in each group.
 The variances of both groups should be equal (use Levene’s Test to check).
CALCULATE THE T-STATISTIC
 Use the formula to compute t.
 Compare the computed t-value with the critical t-value from the t-table.
MAKE A DECISION

If |t-computed| > t-critical, reject H₀ (significant difference exists).

✅ This means the difference between the two group means is statistically significant.
✅ The observed difference is unlikely to be due to random chance alone.
✅ We conclude that there is enough evidence to say that the two groups have different means.

If |t-computed| ≤ t-critical, fail to reject H₀ (no significant difference).

❌ This means the observed difference is not statistically significant.
❌ The difference between the two means could have occurred by chance.
❌ We do not have enough evidence to say that the groups are truly different.
IN CONCLUSION: The independent t-test is powerful for comparing two separate groups to
determine if their means differ significantly. It is widely used in research, business, and
medicine to test hypotheses about group differences.

T TEST FOR DEPENDENT SAMPLE

T TEST FOR DEPENDENT SAMPLE - A dependent (paired) t-test is used to
compare the means of two related groups. It assesses whether the mean difference
between paired observations is statistically significant.
This test is used when the same participants or paired subjects are measured before
and after an intervention or under two different conditions.
When to Use a Dependent (Paired) t-Test?
 Same individuals measured twice (e.g., before vs. after an experiment).

 Matched pairs (e.g., twins, spouses, or students paired by similar

characteristics).

 Repeated measures design (e.g., testing the same group under two
conditions).
OTHER ASSUMPTIONS IN DEPENDENT SAMPLES:
NORMALITY: The differences between paired observations should be approximately
normally distributed.
INDEPENDENCE: The pairs of observations should be independent of each other. This
means that the data points within each pair are related, but the pairs themselves are
independent of each other.
SCALE OF MEASUREMENT: The dependent variable should be measured on a
continuous scale (interval or ratio).
RANDOM SAMPLING: The sample should be randomly selected from the population to
generalize the results properly.
EXAMPLES:
 Effect of Exercise: Testing participants' weight before and after a 3-month
fitness program.
 🏫 Teaching Methods: Comparing students' test scores before and after
implementing a new teaching strategy.
 💊 Medical Study: Measuring blood pressure before and after taking a new
drug.
STEPS TO CONDUCT AN DEPENDENT T-TEST
STATE THE HYPOTHESES

 Null Hypothesis (H₀): There is no significant difference between the means of the two
groups.

 Alternative Hypothesis (H₁): There is a significant difference between the means of the
two groups.
CHECK ASSUMPTIONS
 The data should be independent (from different participants).
 The data should be normally distributed in each group.
 The variances of both groups should be equal (use Levene’s Test to check).
CALCULATE THE T-STATISTIC
 Use the formula to compute t.
 Compare the computed t-value with the critical t-value from the t-table.
MAKE A DECISION

If |t-computed| > t-critical, reject H₀ (significant difference exists).

✅ This means the difference between the two group means is statistically significant.
✅ The observed difference is unlikely to be due to random chance alone.
✅ We conclude that there is enough evidence to say that the two groups have different means.
IN CONCLUSION: The dependent t-test is used when comparing the same subjects under
two conditions. It measures whether the difference in means is significant. It is commonly
used in education, medicine, and psychology.

KEY DIFFERENCES: INDEPENDENT VS. DEPENDENT T-TEST

The Mann-Whitney U test or Wilcoxon sum-rank Test is a nonparametric test that

does not assume normality and instead compares the median ranks of two
independent groups.
The Wilcoxon Signed-Rank Test is a nonparametric test that does not require
normality and instead ranks the differences between paired values
ALTERNATIVE TO PARAMETRICS (NOT NORMALLY DISTRIBUTED)
If the data in an independent samples t-test are not normally distributed, the best
alternative is the Mann-Whitney U test (also called the Wilcoxon rank-sum test).
Why?
 The t-test assumes normality in the distribution of the data. If this assumption is
violated (especially with small sample sizes), results may be unreliable.
 The Mann-Whitney U test is a nonparametric test that does not assume
normality and instead compares the median ranks of two independent groups.

When to use the Mann-Whitney U test?

 When your data are ordinal or not normally distributed.
 When you have small sample sizes where normality can't be assumed.
 When data contain outliers that affect the mean but not the rank-based
comparison.

If the data in a dependent (paired) samples t-test are not normally distributed, the
best alternative is the Wilcoxon Signed-Rank Test.
Why?
 The paired t-test assumes normality of the differences between paired
observations.
 The Wilcoxon Signed-Rank Test is a nonparametric test that does not
require normality and instead ranks the differences between paired values.
When to use the Wilcoxon Signed-Rank Test?
✔ When your paired differences are not normally distributed.
✔ When dealing with ordinal data or small sample sizes.
✔ When data contain outliers that could skew parametric results.
This test compares the median difference instead of the mean, making it more robust
against non-normality and outliers.
If an independent samples t-test violates the assumption of homogeneity of
variance (meaning the two groups have unequal variances), the best alternative is
Welch’s t-test.
Why?
 The standard independent t-test assumes equal variances (homogeneity of
variance).
 Welch’s t-test is a modified version of the independent t-test that does not
assume equal variances, making it more reliable when variances are unequal.
When to use Welch’s t-test?
✔ When Levene’s test (or another variance test) shows significant inequality of
variances.
✔ When sample sizes are unequal, making variance differences more impactful.
✔ When you still meet normality assumptions but violate homogeneity of variance.
It’s basically the best fix when variance inequality is an issue while keeping the power
of the t-test. If normality is also violated, you might need a nonparametric alternative
like the Mann-Whitney U test.

IN CONCLUSION
For Independent T test, if assumptions are violated
 If the data in an independent t-test are NOT normally distributed, I will use the
Mann-Whitney U Test (Wilcoxon Rank-Sum Test).
 If the data in an independent t-test FAILS the homogeneity of variance
assumption (based on Levene’s test), I will use Welch’s t-test.
 If the data is BOTH not normally distributed AND fails homogeneity of variance
(based on Levene’s test), I will use the Mann-Whitney U Test (Wilcoxon Rank-
Sum Test).
For Dependent T test, if assumptions are violated
 If the dependent t-test fails normality → Use Wilcoxon Signed-Rank Test.

 If the dependent t-test fails homogeneity of variance (even though it’s rarely an
issue) Still use Wilcoxon Signed-Rank Test.

 If both normality and homogeneity of variance are violated → Wilcoxon Signed-

Rank Test is the best option.
 ANOVA (DIFFERENT TYPES OF ANOVA)
ANOVA, which stands for ANALYSIS OF VARIANCE, is a statistical test used to analyze the
difference between the means of more than two groups.
ONE WAY ANOVA or ONE WAY ANALYSIS OF VARIANCE is a type of statistical test. It is
used to determine whether there are any statistically significant differences between the
means of three or more independent (unrelated) groups.
OTHER DEFINITIONS: ONE-WAY ANOVA is a statistical test used to determine whether there
are significant differences between the means of three or more independent groups.
WHEN TO USE ONE-WAY ANOVA?
 When you have one independent variable (factor) with three or more groups (e.g.,
types of fertilizers, different teaching methods, etc.).

 When your dependent variable is continuous (numerical) (e.g., test scores, height,
reaction time).

 When your data meet normality and homogeneity of variance assumptions.

EXAMPLE SCENARIO:
Imagine you are testing the effect of three different diets (🍎 Diet A, 🥩 Diet B, 🍕 Diet C) on weight
loss. You measure weight loss (kg) after a month. A One-Way ANOVA will tell you if there is a
statistical difference in mean weight loss between these three diets.

HYPOTHESES IN ONE-WAY ANOVA:

 NULL HYPOTHESIS (H₀): All group means are equal. (No significant difference)

 ALTERNATIVE HYPOTHESIS (H₁): At least one group mean is different.

Key Assumptions of One-Way ANOVA:
 Independence – Each sample is independent of the others.

 Normality – The dependent variable is normally distributed within each group.

 Homogeneity of Variance – Variance among groups should be approximately equal

(Levene’s test can check this).
Interpreting One-Way ANOVA Results:

 If p-value < 0.05 → Reject H₀ (At least one group mean is significantly different).

 If p-value ≥ 0.05 → Fail to reject H₀ (No significant difference).

What If ANOVA is Significant? 🤔
If One-Way ANOVA shows a significant difference, you don’t know which specific groups are
different—only that at least one is. To find out which groups differ, you need post hoc tests
like:
✔ Tukey’s HSD (Honest Significant Difference) – For pairwise comparisons. If variance is
equal
✔ Bonferroni Correction – Adjusts for multiple comparisons. If variance is equal
If the VARIANCE is UNEQUAL in ONE-WAY ANOVA (violating the homogeneity of variance
assumption), the best post hoc test to use is:
GAMES-HOWELL TEST
👉 Why?
 Unlike Tukey’s HSD, Games-Howell does NOT assume equal variances.
 It also works when sample sizes are unequal.
 It adjusts for multiple comparisons, reducing the risk of false positives.
CONCLUSION FOR POST HOC TEST IN ONEWAY ANOVA
 If One-Way ANOVA shows a significant difference among three or more independent
groups, I will use a post hoc test to determine which specific groups are different.
 If the data passes the homogeneity of variance assumption (Levene’s test is not
significant), I will use Tukey’s HSD or Bonferroni Correction.
 If the data fails the homogeneity of variance assumption (Levene’s test is significant), I
will use the Games-Howell Test instead.

ALTERNATIVES TO ONE-WAY ANOVA IF ASSUMPTIONS ARE VIOLATED

If normality holds but VARIANCES ARE UNEQUAL → Use Welch’s ANOVA ✅
If the VARIANCE is UNEQUAL in ONE-WAY ANOVA (violating the homogeneity of variance
assumption), the best post hoc test to use is:
GAMES-HOWELL TEST
👉 Why?
 Unlike Tukey’s HSD, Games-Howell does NOT assume equal variances.
 It also works when sample sizes are unequal.
 It adjusts for multiple comparisons, reducing the risk of false positives.
If normality is also violated → Use Kruskal-Wallis Test ✅
The KRUSKAL-WALLIS TEST is a NONPARAMETRIC ALTERNATIVE to ONE-WAY
ANOVA, used when:
 The data are not normally distributed
  The groups have unequal variances (violating homogeneity of variance).

 The dependent variable is ordinal or continuous but doesn’t meet ANOVA’s

assumptions.
Key Assumptions:
 The samples must be independent.

 The dependent variable should be ordinal or continuous.

 The groups should have the same shape distribution (though variances can be different).
What If Kruskal-Wallis is Significant? 🤔
If KRUSKAL-WALLIS finds a difference, you still don’t know which specific groups are different.
You need post hoc tests like:
✔ Dunn’s Test – Adjusts for multiple comparisons.
✔ Mann-Whitney U Test – For pairwise group comparisons.

CONCLUSION ON POST HOC TEST OF KRUSKAL WALLIS

If the data violates the assumptions of ONE-WAY ANOVA (i.e., homogeneity of variance and
normality of distribution), I will use the KRUSKAL-WALLIS TEST instead.
If Kruskal-Wallis shows a significant difference, I will use post hoc tests to determine which
groups are different:
 Mann-Whitney U Test → For pairwise comparisons between two specific groups.
 Dunn’s Test → For adjusted multiple comparisons across all groups while controlling for
error rate.

TWO-WAY ANOVA or TWO-WAY ANALYSIS OF VARIANCE - It is a statistical test used to

determine the effect of two nominal predictor variables on a continuous outcome variable.
A two-way ANOVA tests the effect of two independent variables on a dependent variable.
OTHER DEFINITION - TWO-WAY ANOVA (ANALYSIS OF VARIANCE) is a statistical test
used to compare the means of a dependent variable across groups formed by two
independent categorical variables (factors). It helps determine:
 If each factor has a significant effect on the dependent variable (Main Effects).

 If the two factors interact and influence the dependent variable together (Interaction
Effect).
WHEN TO USE TWO-WAY ANOVA?
  You have two independent categorical variables (factors).

 You have one continuous dependent variable (e.g., test scores, height, sales).

 You want to analyze both individual and combined effects of the factors.

 The data meets normality and homogeneity of variance assumptions.

Imagine you’re studying how teaching method and gender affect students' exam scores.

 Factor 1 (Teaching Method) → Traditional vs. Online vs. Hybrid

 Factor 2 (Gender) → Male vs. Female
 Dependent Variable → Exam Scores

A Two-Way ANOVA will help you analyze:

✔ Main Effect of Teaching Method (Does the method affect scores?)
✔ Main Effect of Gender (Do males and females score differently?)
✔ Interaction Effect (Does the effect of teaching method depend on gender?)

HYPOTHESES IN TWO-WAY ANOVA:

 H₀ (Null Hypothesis): No significant differences exist.

o H₀₁: The first factor (e.g., teaching method) has no effect.

o H₀₂: The second factor (e.g., gender) has no effect.

o H₀₃: There is no interaction effect between the two factors.

 H₁ (Alternative Hypothesis): At least one factor or their interaction is significant.

KEY ASSUMPTIONS OF TWO-WAY ANOVA:

 INDEPENDENCE – Each observation is independent of the others.

 NORMALITY – The dependent variable is normally distributed within each group.

 HOMOGENEITY OF VARIANCE – Variance should be approximately equal across

groups (Levene’s test checks this).

INTERPRETING TWO-WAY ANOVA RESULTS:

✔ SIGNIFICANT MAIN EFFECTS → If a factor (e.g., teaching method) is significant, it
independently influences the dependent variable.
✔ SIGNIFICANT INTERACTION EFFECT → If there’s a significant interaction, the effect of one
factor depends on the level of the other.
✔ POST HOC TESTS (e.g., Tukey’s HSD) → If significant, post hoc tests help determine
which specific groups differ.

WHAT IF ASSUMPTIONS ARE VIOLATED?

If Two-Way ANOVA assumptions are violated (i.e., variance is not homogeneous & distribution
is not normal), you should use nonparametric alternatives like:
SCHEIRER-RAY-HARE TEST (BEST ALTERNATIVE FOR TWO-WAY ANOVA)
 Non-parametric equivalent of Two-Way ANOVA

 Works with independent (between-group) factors

 Uses ranked data like Kruskal-Wallis

ALIGNED RANK TRANSFORM (ART) ANOVA
 Best for non-normal data with heterogeneous variances.

 Transforms data into ranks, then runs a standard ANOVA on the transformed data.

 Works even when there’s an interaction effect between factors.

REPEATED MEASURES ANOVA (Analysis of Variance) is a statistical test used to compare

the means of three or more related groups. It is commonly applied when the same subjects
are measured multiple times under different conditions or over time.
KEY FEATURES OF REPEATED MEASURES ANOVA:
WITHIN-SUBJECTS DESIGN – The same participants are tested multiple times rather than
having different groups.
REDUCES VARIABILITY – Since each subject serves as their own control, individual
differences are minimized, increasing statistical power.
COMPARES MULTIPLE CONDITIONS OR TIME POINTS – It is used in experiments where
participants undergo different treatments or are tested at different time intervals.
COMMON APPLICATIONS:
Measuring the effect of a drug over different time points (e.g., before, during, and after
treatment).
Comparing student test scores across different teaching methods.
Evaluating psychological responses to various stimuli.
ASSUMPTIONS OF REPEATED MEASURES ANOVA:
NORMALITY – The data should be normally distributed.
SPHERICITY – The variances of differences between conditions should be approximately equal
(tested using Mauchly’s test).
If the data failed in Mauchly’s test, we use Greenhouse Geisser for extreme violation in
sphericity and Huynh- Feldt is used for mild violation in sphericity.
NO EXTREME OUTLIERS – Outliers can impact the results significantly.
Example Scenario:
Imagine a psychology experiment testing the effect of sleep deprivation on cognitive
performance. A group of participants takes a memory test after:
8 hours of sleep
4 hours of sleep
No sleep
NON-PARAMETRIC STATISTICS FOR VIOLATING ASSUMPTIONS
Friedman Test – The most common non-parametric alternative.
 Used when you have three or more related samples.
 Works like a rank-based version of repeated measures ANOVA.
 Example: Testing students' test scores under three different teaching methods.

F MIXED-METHOD ANOVA - known a Mixed Design ANOVA or Split-Plot ANOVA, is a

powerful statistical technique used to analyze data involving both between-subjects and
within-subjects factors.
OTHER DEFINITIONS - Mixed-Methods ANOVA is a type of ANOVA that combines both
within-subjects and between-subjects factors in one analysis.
🔹 Within-Subjects Factor → The same participants are measured multiple times (e.g., before
and after treatment).
🔹 Between-Subjects Factor → Different groups are compared (e.g., males vs. females).
Example Scenario:
A researcher studies the effect of exercise intensity (Low, Medium, High) on reaction time,
while also considering gender (Male, Female).
 Exercise intensity = Within-Subjects Factor (same participants tested under all
conditions).
 Gender = Between-Subjects Factor (comparison between male & female groups).

KEY FEATURES OF MIXED-DESIGN ANOVA:

 Allows for both repeated measures and independent groups.

 Helps analyze interaction effects between factors.

 Requires checking sphericity (for within-subject factors) and homogeneity of

variance (for between-subject factors).

WHEN TO USE MIXED-DESIGN ANOVA?

 Studying longitudinal effects (e.g., performance over time across different groups).
 Examining how individual differences (e.g., gender, age) impact repeated measures.
 Experiments where some factors are measured repeatedly, while others involve
separate groups.
NON-PARAMETRIC ALTERNATIVE FOR MIXED-DESIGN ANOVA
1. Aligned Rank Transformation (ART) ANOVA
 Non-parametric alternative to Mixed ANOVA.
 Uses rank transformation while still allowing factorial analysis (like interactions).
 Handles both within- and between-subjects factors properly.
 Best when sample size is small or normality is violated.

ANALYSIS OF COVARIANCE (ANCOVA) - is a statistical technique that combines the feature

of Analysis of Variance (ANOVA) and regression analysis.
OTHER DEFINITION: ANCOVA is a statistical test that compares the means of different
groups (Independent) while controlling for the effects of other continuous variables
(covariates).
KEY FEATURES OF ANCOVA:
 Combination of ANOVA & Regression – Like ANOVA, it compares means, but like
regression, it adjusts for confounding variables.

 Removes Unwanted Variability – By controlling for covariates, ANCOVA increases

statistical power.

 Tests for Group Differences More Accurately – Adjusts group means by accounting for a
covariate's influence.
EXAMPLE SCENARIO:
A researcher wants to study the effect of teaching method (Lecture, Group Work, Online
Learning) on final exam scores, but some students started with different IQ levels.
 INDEPENDENT VARIABLE (IV): Teaching Method (Categorical: 3 groups).
 DEPENDENT VARIABLE (DV): Final Exam Scores (Continuous).
 COVARIATE: IQ Score (Continuous).
ANCOVA helps remove the influence of IQ so that the true effect of teaching method is
clearer.
ASSUMPTIONS OF ANCOVA:
LINEARITY – The relationship between the covariate and DV must be linear. (Use Scatter
Plot)
HOMOGENEITY OF REGRESSION SLOPES – The effect of the covariate should be similar
across all groups. (Check the regression coefficient)
INDEPENDENCE & NORMALITY – Residuals should be normally distributed. (Shapiro-Wilk
Test, Q-Q Plot for better representation)
HOMOSCEDASTICITY – Equal variance across groups. (Levene’s Test)

WHEN TO USE ANCOVA?

 When comparing group means while controlling for external influences.

 When a covariate might affect the dependent variable and needs to be adjusted for.

 When pre-test scores are used to control for baseline differences (e.g., in educational
research).

QUADE’S TEST (BEST ALTERNATIVE TO ANCOVA)

 Controls for covariates while testing differences between groups.
  Works for ordinal & non-normally distributed data.

 More powerful than rank-based alternatives.

CORRELATIONAL STATISTICAL METHODS

PEARSON’S R CORRELATION - Pearson’s correlation is one of the most used statistical
methods. It is a suitable method when you have two variables and you want to know if there is
a certain relationship, direction, as well as strength.
OTHER DEFINITIONS: Pearson's r (also known as the Pearson correlation coefficient) is a
statistical measure that describes the strength and direction of the linear relationship
between two continuous variables. It’s one of the most common ways to assess how strongly
two variables are related.
Key Points:
 Range of Values:
Pearson’s r ranges from -1 to +1:
o +1: Perfect positive correlation (as one variable increases, the other also
increases perfectly).
o -1: Perfect negative correlation (as one variable increases, the other decreases
perfectly).
o 0: No correlation (no linear relationship between the variables).

Interpretation of Pearson’s r:
 +0.70 to +1.00: Strong positive correlation.
 +0.30 to +0.69: Moderate positive correlation.
 +0.01 to +0.29: Weak positive correlation.
 0.00: No linear correlation.
 -0.01 to -0.29: Weak negative correlation.
 -0.30 to -0.69: Moderate negative correlation.
 -0.70 to -1.00: Strong negative correlation.

ASSUMPTIONS FOR PEARSON'S R:

1. Linear Relationship: The relationship between the two variables should be linear.
(Scatter Plot)
 2. Normality: Both variables should be approximately normally distributed. (Q-Q Plot ,
Shapiro-Wilks test)
3. Homogeneity of Variance: The variability in one variable should be similar across all
values of the other variable. (Scatter Plot, Residual plot, Levenes test)

HOW TO INTERPRET PEARSON'S R:

 A strong positive r (close to +1) means that as one variable increases, the other tends to
increase as well.
 A strong negative r (close to -1) means that as one variable increases, the other tends to
decrease.
 A r near 0 means no linear relationship exists.

EXAMPLE:
If you measure the height and weight of a group of people and compute Pearson’s r, a result of
+0.85 suggests that there’s a strong positive correlation—as height increases, weight tends to
increase too.

NON-PARAMETRICS ALTERNATIVE (IF ASSUMPTIONS ARE NOT SATISIFIED)

SPEARMAN'S RANK CORRELATION (SPEARMAN’S RHO)
 Non-parametric alternative to Pearson’s r.

 Measures the strength and direction of the monotonic relationship between two
variables.

 Works with ordinal data or continuous data that isn’t normally distributed.

 Can handle non-linear relationships as long as the relationship is monotonic (either

always increasing or always decreasing).
When to Use:
 When the variables are not normally distributed.
 When the relationship between the variables is monotonic but not necessarily linear.
 When you have ordinal data.

Kendall’s Tau (τ)

  Another non-parametric test that measures the strength and direction of the monotonic
relationship between two variables.

 It’s more robust than Spearman’s rho for smaller datasets or tied ranks.
When to Use:
 Similar to Spearman’s rho, but especially when you have small sample sizes or many
tied ranks.

MULTIPLE CORRELATION - is the study of combined influence of two or more variables

on a single variable. If we have more than two variables which are interrelated in some way
and our interest is to know the relationship between one variable and a set of others, we use
Multiple Correlation.
OTHER DEFINITION - Multiple correlation (denoted as R) measures the strength of the
relationship between one dependent variable (Y) and two or more independent variables
(X₁, X₂, X₃, ...). It is commonly used in multiple regression analysis to assess how well a set
of predictor variables collectively explain the variation in the dependent variable.
KEY POINTS ABOUT MULTIPLE CORRELATION (R):
✔ Symbol: R (not to be confused with Pearson’s r)
✔ Range: 0 to 1
✔ Interpretation:
 R = 1: Perfect positive correlation (strong relationship)
 R = 0: No correlation (predictors have no effect)
 Higher values of R indicate a stronger collective relationship between predictors
and the dependent variable.
EXAMPLE OF MULTIPLE CORRELATION (R) IN REGRESSION:
A researcher wants to predict college GPA (Y) based on:

 High school GPA (X₁)

 SAT scores (X₂)

 Hours studied per week (X₃)

After running a multiple regression analysis, they obtain R = 0.85.
🔹 Interpretation: 85% of the variation in college GPA can be explained by the combined effect of
high school GPA, SAT scores, and study hours.
ASSUMPTIONS IN MULTIPLE CORRELATION ANALYSIS:
1. Linear Relationship: The relationship between the two variables should be linear.
(Scatter Plot)
2. Normality: Both variables should be approximately normally distributed. (Q-Q Plot ,
Shapiro-Wilks test)
3. Homogeneity of Variance: The variability in one variable should be similar across all
values of the other variable. (Scatter Plot, Residual plot, Levenes test)

SPEARMAN'S RANK CORRELATION (SPEARMAN’S RHO)

 Non-parametric alternative to Pearson’s r.

 Measures the strength and direction of the monotonic relationship between two
variables.

 Works with ordinal data or continuous data that isn’t normally distributed.

 Can handle non-linear relationships as long as the relationship is monotonic (either

always increasing or always decreasing).
When to Use:
 When the variables are not normally distributed.
 When the relationship between the variables is monotonic but not necessarily linear.
 When you have ordinal data.

Kendall’s Tau (τ)

 Another non-parametric test that measures the strength and direction of the monotonic
relationship between two variables.

 It’s more robust than Spearman’s rho for smaller datasets or tied ranks.
When to Use:
 Similar to Spearman’s rho, but especially when you have small sample sizes or many
tied ranks.

PARTIAL CORRELATION - Partial correlation is a measure of the strength and direction of

a linear relationship between two continuous variables whilst controlling for the effect of
one or more other continuous variables (also known as 'covariates' or 'control' variables).
Although partial correlation does not make the distinction between independent and
dependent variables. Partial correlation coefficient is a coefficient to describe the relationship
between X and Y when taking away the effects of control variable Z, which can be used to test
conditional independence. (Parang katulad sya ng Ancova)
OTHER DEFINITION: Partial correlation measures the relationship between two variables
while controlling the effect of one or more other variables. It helps determine whether two
variables are truly related or if their correlation is due to a third (confounding) variable.
ASSUMPTIONS IN MULTIPLE CORRELATION ANALYSIS:
1. Linear Relationship: The relationship between the two variables should be linear.
(Scatter Plot)
2. Normality: Both variables should be approximately normally distributed. (Q-Q Plot ,
Shapiro-Wilks test)
3. Homogeneity of Variance: The variability in one variable should be similar across all
values of the other variable. (Scatter Plot, Residual plot, Levenes test)
4. Covariate(s) - You should only perform partial correlation if you have one or more
covariates. A covariate is a variable whose affects you want to remove when examining
the variable relationship of interest.
EXAMPLE:
A researcher wants to study the correlation between study hours (X) and exam scores (Y) but
controls for IQ level (Z) since IQ might influence both study hours and exam scores.
 A simple Pearson’s correlation between study hours and exam scores might be
misleading because IQ could be a confounding factor.
 Using partial correlation, the researcher can isolate the effect of study hours on exam
scores while eliminating the influence of IQ.
HOW TO INTERPRET PARTIAL CORRELATION

 rXY⋅Z>0r_{XY \cdot Z} > 0rXY⋅Z>0 → Positive relationship between X and Y even after
controlling for Z.

 rXY⋅Z<0r_{XY \cdot Z} < 0rXY⋅Z<0 → Negative relationship between X and Y after

controlling for Z.

 rXY⋅Z≈0r_{XY \cdot Z} \approx 0rXY⋅Z≈0 → No significant relationship after controlling

for Z.

NON-PARAMETRIC ALTERNATIVE IF ASSUMPTIONS ARE VIOLATED

SPEARMAN’S RANK PARTIAL CORRELATION (NON-PARAMETRIC ALTERNATIVE TO
PEARSON’S PARTIAL CORRELATION)
 Use when: Data is not normally distributed or ordinal.
 How it works:
 o First, rank all values.

o Then, compute the partial correlation formula but using Spearman’s rank
correlation (ρ) instead of Pearson’s correlation (r).
 Interpretation: Similar to Pearson’s partial correlation, but based on ranked data.
KENDALL’S TAU PARTIAL CORRELATION (FOR SMALL SAMPLE SIZES)
 Use when:
o Non-normality + small sample size (n < 20).

o Many tied ranks in the data.

 Why use it? More robust than Spearman when the data contains ties or outliers.

REGRESSION ANALYSIS - This is a statistical tool that is used to predict trends by

examining the relationship between two or more variables. Simple Linear regression is
used to predict the relationship between two variables, while Multiple Linear regression is used
for more than two variables.
OTHER DEFINITION: REGRESSION ANALYSIS is a statistical technique used to model the
relationship between one or more independent variables (predictors) and a dependent variable
(outcome). It helps to:
 Predict values of the dependent variable.

 Determine relationships between variables.

 Assess the strength & significance of predictors.

 Control for confounding variables.

TYPES OF REGRESSION ANALYSIS

1. SIMPLE LINEAR REGRESSION
 Used when: One independent variable (X) predicts a continuous dependent variable (Y).
 Example: Predicting exam scores (Y) based on study hours (X).
2. MULTIPLE LINEAR REGRESSION
 Used when: More than one independent variable predicts a continuous dependent
variable.

 Example: Predicting house price (Y) based on square footage (X₁), number of rooms
(X₂), and location (X₃).
ASSUMPTIONS OF REGRESSION ANALYSIS

 LINEARITY – The relationship between X and Y is straight. (Scatter plot, Residual Plot)

 NORMALITY OF RESIDUALS – Errors (ε) are normally distributed. (Histogram, Q-Q

Plot)

 HOMOSCEDASTICITY – Variance of errors remains constant. (Residual Plot)

 NO MULTICOLLINEARITY – Predictors should not be highly correlated.

 INDEPENDENCE OF ERRORS – Observations should not be correlated. (Durbin

Watson Test)

NON PARAMETRICS TEST USED IN REGRESSIONAL ANALYSIS

SPEARMAN’S RANK CORRELATION:
 Used when you want to assess the monotonic relationship between two variables, rather
than a linear one. It’s useful for ordinal data.
KENDALL’S THEIL REGRESSION (ALSO KNOWN AS THEIL-SEN ESTIMATOR)
 is a nonparametric alternative to linear regression, but it’s not necessarily
the best or only choice for all situations.

OTHER NON-PARAMETRIC TEST

The CHI-SQUARE TEST OF INDEPENDENCE is a statistical method used to determine
whether there is a significant association between two categorical variables. This non-
parametric test is particularly useful in analyzing data in contingency tables, where the
variables are divided into categories. The test compares the observed data with a
hypothetical scenario where no relationship exists between the variables.
OTHER DEFINITIONS- The Chi-Square (χ²) test is a non-parametric statistical test used to
determine if there is a significant association between categorical variables. It checks
whether the observed frequencies differ from expected frequencies by chance.
 Determine Relationships
  Support-Decision Making
 Test Hypotheses
ASSUMPTIONS
 Independence of observed data
 Sample size must be large (If small consider using the fisher exact test)
 Frequency must be 5 above or else the result will not be valid
LIMITATIONS:
 Data Type
 Sample Size Dependence
 Expected Frequency Requirement
 Assumes Independence
 No Measure of Strength
When to Use Chi-Square?

 Categorical data (nominal or ordinal)

 Large sample size (expected counts in each cell ≥ 5)

 Independence of observations

IN CONCLUSION: You reject the null hypothesis in a Chi-Square test when the p-value is
smaller than the chosen significance level (usually 0.05). This indicates that the observed
frequencies are significantly different from what you would expect under the null hypothesis,
implying a relationship or association between the variables.

MCNEMAR TEST - A statistical test used to compare paired nominal data to assess if there
is a significant difference between two related groups. Evaluates changes or effects in
dichotomous (binary) outcomes, like Yes/No or Success/Failure. Utilized in fields like
psychology, medicine, and education to determine the impact of interventions, therapies, or
treatments.
OTHER DEFINITION: The McNemar Test is a non-parametric test used to analyze paired
nominal data, specifically when you have two related dichotomous variables. It tests
whether there is a significant difference in the proportion of cases between two related groups
(before and after treatment, or two different conditions).
NULL HYPOTHESIS:
There is no significant difference between the two related groups.
The changes in responses are symmetrical.
Example: The number of participants who improve is equal to those who
worsen.
ALTERNATIVE HYPOTHESIS:
If p < 0.05, there is a significant difference between the two related
groups.
The changes in responses are asymmetric.
Example: More participants improve than worsen after treatment.
ASSUMPTIONS OF MCNEMAR:
 Paired data (before/after, treatment/control, etc.).
 Binary data (two categories).
 Symmetry of categories.
 Sufficient sample size for each outcome category (at least 5).
 Independence of pairs.
 The test detects binary changes between two conditions.
When to Use McNemar's Test?
1. Paired Data: You have paired observations (same subjects tested twice or matched
groups).
o Example: A pre-test/post-test design (e.g., testing opinions on a product before
and after using it).
2. Two Dichotomous Variables: You are working with two categorical variables that have
two categories each (Yes/No, Success/Failure, etc.).
SUMMARY:
 McNemar's Test is used to analyze paired nominal data.
 It compares the proportion of changes between two conditions (before and after).
 It is commonly used in pre-test/post-test designs.
 The test statistic is based on the difference between the "failure to success" and
"success to failure" pairs.
 If p-value < 0.05, you reject the null hypothesis, indicating a significant difference
between the two conditions.

RISK RATIO (RR), also called RELATIVE RISK (RR), it is non-parametric where it measure
used in epidemiology and medical research to compare the probability (risk) of an event
occurring in two different groups.
It is commonly used in cohort studies to determine if exposure to a certain factor (e.g.,
smoking, medication, lifestyle) increases or decreases the risk of an outcome (e.g., disease,
recovery, death).
ASSUMPTIONS:
  Proper Study Design (Cohort or Prospective Study)
 Dichotomous (Binary) Outcome
 Independence of Observations
 Adequate Sample Size
 No Confounding Variables (or Controlled for Confounders)
INTERPRETING RISK RATIO (RR)
 RR = 1 → No difference in risk (exposure has no effect).
 RR > 1 → Exposure increases risk (e.g., smoking increases the risk of lung cancer).
 RR < 1 → Exposure decreases risk (e.g., a vaccine reduces the risk of infection).

SUMMARY:
 Risk Ratio (RR) compares the risk of an event occurring in an exposed group vs. an
unexposed group.
 Used in cohort studies to measure relative risk.
 RR = 1 → No effect, RR > 1 → Increased risk, RR < 1 → Reduced risk.
 RR is different from Odds Ratio (OR) and is preferred for prospective studies.

KENDALL'S COEFFICIENT OF CONCORDANCE, is also known as Kendall's W, is a non-

parametric statistical measure used to assess the agreement or concordance among a
group of judges or raters of (n) objects.
It is a normalization of the friedman test. The value of W ranges from 0 to 1, with 0
indicating no agreement and 1 indicating complete agreement.
OTHER DEFINITIONS: Kendall’s W is a non-parametric statistic that measures the degree
of agreement (concordance) between multiple raters or judges who rank a set of items. It
ranges from 0 to 1, where:
 W = 1 → Perfect agreement among raters
 W = 0 → No agreement (ratings are completely random)
 0 < W < 1 → Partial agreement
IT IS COMMONLY USED IN:
 Inter-rater reliability (e.g., multiple judges ranking contestants in a contest)

 Survey research (e.g., agreement on importance of factors)

 Behavioral sciences & psychology

WHEN TO USE KENDALL’S W?
  When you have ordinal (ranked) data.

 When 3 or more raters/judges rank multiple items.

 When normality assumptions are violated (since it’s non-parametric).

Interpreting Kendall’s W
 W = 1.0 → Perfect agreement among raters.
 W = 0.7 - 0.9 → Strong agreement.
 W = 0.3 - 0.6 → Moderate agreement.
 W < 0.3 → Weak agreement (raters mostly disagree).
If W is close to 0, it means the raters have different opinions, and there is low reliability.
IN CONCLUSION:
 Kendall’s W measures agreement among multiple raters.
 Used for ordinal (ranked) data and more than two raters.

 Non-parametric (no normality assumption needed).

 Higher W means stronger agreement.

CRONBACH’S ALPHA - also known as tau-equivalent reliability or coefficient alpha It was

named after the American psychologist Lee Cronbach. Is a statistical measure used to assess
the internal consistency, or reliability, of a set of items (usually survey or test questions) that
are intended to measure the same underlying construct or latent variable.
ASSUMPTIONS:
 Unidimensionality
 (tau equivalent) equal item contribution
 sufficient items
 no extreme correlations. (Multicollinearity)
WHAT IF CRONBACH’S ALPHA IS LOW?

 Remove weak items (check item-total correlations).

 Reword or replace ambiguous questions.

 Ensure all items measure the same construct.

 Use McDonald’s Omega (ω) instead, which is a better alternative when tau-
equivalence is violated.

SUMMARY
  Cronbach’s Alpha measures internal consistency (reliability).

 It is used to check if test items are consistently measuring the same

construct.

 Higher α = More reliable scale, but extremely high values (≥ 0.95) may indicate
redundancy.

 Alternatives: McDonald’s Omega (ω) is more robust in some cases.

TRANSFORMING SCORES - Score transformations are used to convert raw scores into
transformed scores that can be compared and interpreted. The purpose of transforming scores
is to give meaning to the scores and to allow for direct comparison between scores.

EXPLORATORY FACTOR ANALYSIS (EFA) – It is a non-parametric statistical method that

identifies underlying relationships between large number of variables by grouping them
into fewer groups that are manageable “Factors”. It is a data reduction technique that tries
to simplify complex data by explaining the variance among the observed variables
(Factors) with a smaller number of latent or unobserved variables.
OTHER DEFINITION – Exploratory Factor Analysis (EFA) is a statistical technique used to
identify the underlying structure in a set of observed variables. It helps researchers reduce
data dimensions by grouping correlated variables into latent factors that explain common
variance.

WHEN TO USE EFA?

 When you don’t know the exact number of factors in your data.

 When you want to reduce the number of variables while keeping meaningful
information.

 When you need to check if survey items measure the same construct (e.g., personality
traits, psychological scales).

CHECK SUITABILITY OF DATA FOR FACTOR ANALYSIS

 Bartlett’s Test of Sphericity → Checks if correlations between variables are significant.

 Kaiser-Meyer-Olkin (KMO) Test → Measures sample adequacy (should be ≥ 0.60).

INTERPRET FACTOR LOADINGS & NAME THE FACTORS
 Factor loadings ≥ 0.40 → Meaningful relationship between item & factor.
 Variables with high cross-loadings → Consider removing or modifying.
 Naming factors → Based on grouped variables (e.g., "Emotional Stability", "Cognitive
Ability").

SUMMARY

 Exploratory Factor Analysis (EFA) identifies latent variables in a dataset.

 Helps in data reduction & identifying underlying structures.

 Requires KMO test, Bartlett’s test, factor extraction, rotation, and interpretation.

 Used in psychology, social sciences, and survey research.

MEDIATION ANALYSIS - Statistical mediation analysis refers to statistical procedures

aimed at testing the hypothesis that an independent variable (X) affects an outcome (Y)
through one or more mediators often presented as (Z).
OTHER DEFINITION – Mediation analysis is a statistical technique used to determine whether
the relationship between an independent variable (X) and a dependent variable (Y) is
influenced by a mediator variable (M).
BARON AND KENNY METHOD - The Baron and Kenny (1986) method is an analysis
strategy for testing mediation hypotheses. In this method for mediation, there are two paths
to the dependent variable. The independent variable must predict the dependent variable,
and the independent variable must predict the mediator.
MEDIATION MODEL STRUCTURE
Direct Effect (c') → The direct relationship between X and Y without the mediator.
Indirect Effect (a × b) → The effect of X on Y through the mediator M.
Total Effect (c) → The sum of the direct and indirect effects.
WHEN TO USE MEDIATION ANALYSIS?
 To check if X influences Y through an intermediate process (M).

 To explore psychological mechanisms in behavioral studies.

 To test causal pathways in experimental & survey research.

TYPES OF MEDIATION
 Full Mediation – When X no longer affects Y after accounting for M (c' ≈ 0).

 Partial Mediation – When X still affects Y, but M significantly reduces this effect.
  No Mediation – If M does not significantly influence Y or if indirect effect (a × b) is not
significant.
SUMMARY
 Mediation analysis checks if an independent variable (X) affects a dependent
variable (Y) through a mediator (M).

 Direct effect (c') and indirect effect (a × b) are tested to determine mediation strength.

 Uses Sobel test, bootstrapping, or structural equation modeling (SEM).

 Helps in psychological, behavioral, and social science research.