MODULE 4: Inferential statistics

4.1 Steps in hypothesis testing; Calculation of z scores

4.2 Parametric and non-parametric tests
4.3 Independent and dependent t tests
4.4 Pearson’s correlation and Regression
4.5 ANOVA and MANOVA

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.1 Steps in hypothesis testing Cont’d..

Hypothesis testing means to tell (on the basis of the data the 2. Selecting a significance level: the hypothesis is tested on a
researcher has collected) whether or not the hypothesis seems pre-determined level of significance and it must be specified.
to be valid. Whether it needs to be accepted or rejected. Generally in practice 5% level or 1% level is adopted for this
1. Setting up the hypotheses: the formulation of hypotheses is purpose. Factors that affects the level of significance are size of
an important step which must be accomplished with due care the samples, variability of measurements, whether the
in accordance with the object and nature of the problem under hypothesis is directional or non-directional and the difference
consideration. It also indicates whether we should use one- between the sample means. The level of the significance must
tailed (directional hypothesis) or two-tailed test (null be adequate in the context of the purpose and nature of
hypothesis or non-directional hypothesis). enquiry.
The following are the possibilities of formulating hypothesis.
• HO: Coffee has no effect on attention span (μ₁ = μ₂).
• H1: Coffee affects attention span (μ₁ ≠ μ₂).
• H1: coffee increases attention span (μ₁ > μ₂) (mean score of
attention span of experimental group > mean score of
attention span of control group)

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.1 Steps in hypothesis testing Cont’d.. EXTRA
 The level of significance (LOS) sets the threshold for rejecting the null  Two-tailed tests (Non-directional): These test for effects in both directions.
hypothesis (H0H_0H0​). It is typically set at 0.05 (5%) or 0.01 (1%), depending They are more conservative, splitting the LOS across two tails (e.g., 0.025
on the study's requirements. The choice of LOS depends on several factors: per tail for α=0.05), making it harder to reject H0​.
1. Sample Size 4. Variability in Data
 Larger samples provide more reliable estimates, reducing variability and  High variability increases uncertainty, making it harder to detect true
increasing the test's power. This allows researchers to detect small effects. A stricter LOS (e.g., 0.01) is used to avoid false positives.
differences and use stricter LOS (e.g., 0.01).  Low variability improves precision, allowing for a more relaxed LOS (e.g.,
 Smaller samples are more variable, making it harder to detect effects. A 0.05).
less stringent LOS (e.g., 0.05) is often preferred to avoid missing true 5. Magnitude of Differences Between Groups
effects.  Small differences between groups are harder to detect, so stricter LOS
[Link] Size (e.g., 0.01) is needed to ensure only strong evidence rejects H0H_0H0​.
 Larger effect sizes (e.g., strong correlations or big group differences) make  Larger differences make detecting effects easier, allowing for a higher LOS
it easier to detect significance. In such cases, researchers may use a higher (e.g., 0.05).
LOS (e.g., 0.05).
 Smaller effect sizes require stronger evidence to detect, so a stricter LOS Balancing Type I and Type II Errors
(e.g., 0.01) is often chosen to minimize false positives.  Type I Error (α): Rejecting H0​ when it is true. A smaller LOS (e.g., 0.01)
3. Directionality of Hypothesis reduces this risk.
 One-tailed tests (Directional): These specify the direction of the effect (e.g.,  Type II Error (β): Failing to reject H0​ when it is false. A larger LOS (e.g.,
"X increases Y"). They are more powerful, so researchers may use a stricter 0.05) reduces this risk.
LOS (e.g., 0.01).  Researchers choose LOS based on the study's goals, consequences of
errors, and data characteristics.
Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai
 4.1 Steps in hypothesis testing Cont’d..
3. Test statistics: and whether the test is one-tailed or two-tailed.
• In simple terms, a test statistic is a number we calculate from the • The critical value divides the test statistic's distribution into two
data we collect in a study. It helps us check if what we observed in regions: the rejection region, where H0 is rejected, and the non-
the sample fits with what we expected based on the null rejection region, where H0 is not rejected.
hypothesis (which assumes no effect or no difference). • For example, in a Z-test with α=0.05, the critical values for a two-
• We compare this test statistic to a critical value (a threshold) or tailed test are −1.96 and +1.96. If the test statistic falls outside
use it to find a p-value (a probability). This comparison tells us these values, it indicates the result is statistically significant,
whether to reject the null hypothesis or not. leading to the rejection of H0.
• The researcher picks the right type of test based on the kind of
data, the size of the sample, and the question they are trying to 5. Decision:
answer. Comparing the value of test statistics and critical value, we make the
• We calculate a test statistic from our data, which tells us how far decision about rejecting or not rejecting the null hypothesis.
our results are from what we'd expect under the null hypothesis.
This statistic is plotted on the x-axis of the normal curve.

4. Critical Value:
• A critical value is a threshold in hypothesis testing that helps
decide whether to reject the null hypothesis. It is determined using
the distribution of the test statistic, the prefix significance level (α),
Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai
 4.1 Steps in hypothesis testing Cont’d..

We reject the null hypothesis when:

a. Value of test statistic < lower critical value OR value of
test statistic > Upper critical value (in case of two-tailed
test)

b. Value of test statistic > critical value (in case of right-

tailed test)

c. Value of test statistic < critical value (in case of left-test)

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.1 Calculation of z score
 When the researcher wants to test the significance of difference 2. Two-sample Z-test: Used to compare the means of two
between the two means, he/she uses either t-test or z test independent samples taken from a normally distributed
 z-test is a parametric test and can be used when the sample size populations.
is more than 30 to test the significant difference between two
means.
 Assumptions of Z-test
 The sample is randomly selected from the population.
 The population is normally distributed or the sample size is large enough (n
3. Paired Z-test: Used to compare the means of two related
> 30) to assume normality.
 The population standard deviation is known. samples (e.g., before-and-after observations).

Types of z test: 4. Z test for proportions: A Z-test for proportions is a statistical

1. One sample z test: Used to compare a sample mean to a test used to determine whether a sample proportion is
known population mean. In this case the population must be significantly different from a known population proportion.
normally distributed and standard deviation is known.

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.1 Calculation of z score Cont’d…

Exercise: One sample z-test

a. A company claims that the average height of its employees is 175 cm. A random sample of 36 employees has a
mean height of 172 cm. If the population standard deviation is 5 cm, test the hypothesis that the mean height of the
company's employees is 175 cm. Use a significance level of 0.05.
b. A manufacturing company claims that the average weight of its products is 500 grams. A random sample of 64
products has a mean weight of 495 grams. If the population standard deviation is 10 grams, test the hypothesis that
the mean weight of the products is 500 grams. Use a significance level of 0.05.
c. A school claims that the average GPA of its students is 3.0. A random sample of 36 students has a mean GPA of 2.8.
If the population standard deviation is 0.5, test the hypothesis that the mean GPA of the students is 3.0. Use a
significance level of 0.05.

CRITICAL VALUES FOR TWO –TAILED Z TEST

Alpha = 0.05 Z = ±1.96

Alpha = 0.01 Z = ±2.58

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.1 Calculation of z score Cont’d…
Exercise: Two sample z-test for independent groups

a. A researcher wants to compare the average heights of men and women in a population. A random sample of 40 men has a mean height of

175 cm, and a random sample of 40 women has a mean height of 160 cm. If the population standard deviation for men is 5 cm and for women

is 4 cm, test the hypothesis that the mean heights of men and women are equal. Use a significance level of 0.05.

b. A survey wants to compare the average amount of time spent exercising per week by people who live in urban and rural areas. A random

sample of 64 people from urban areas has a mean exercise time of 3 hours, and a random sample of 64 people from rural areas has a mean

exercise time of 2 hours. If the population standard deviation for urban areas is 1 hour and for rural areas is 0.8 hours, test the hypothesis that

the mean exercise times of people in urban and rural areas are equal. Use a significance level of 0.01.

c. A researcher wants to compare the average scores of students from two different schools on a standardized test. A random sample of 36

students from School A has a mean score of 80, and a random sample of 36 students from School B has a mean score of 85. If the population

standard deviation for School A is 5 and for School B is 6, test the hypothesis that the mean scores of students from the two schools are

equal. Use a significance level of 0.05. CRITICAL VALUES FOR TWO –TAILED Z TEST

Alpha = 0.05 Z = ±1.96

Alpha = 0.01 Z = ±2.58

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.2 Parametric and Non-parametric tests

Parametric tests are statistical tests that require specific assumptions normality can be tested using statistical software.
about the data being analyzed. These assumptions primarily involve: • Statistical Power: They are preferred over nonparametric tests
• Random Sampling: The data must be randomly selected from the because they are more efficient and precise when assumptions are
population. met.
• Independence: Observations within the sample must not influence
each other. Examples of Parametric Tests
• Normality: The population from which the sample is drawn should • t-Test: Used to compare the means of two groups or a sample mean
follow a normal distribution (or approximately normal). against a known value. Assumes normality and equal variances (in
• Homogeneity of Variance: For certain tests, the variances of groups independent samples).
being compared should be roughly equal. • Analysis of Variance (ANOVA): Compares means across three or
When these assumptions are met, parametric tests are more statistically more groups. Assumes normal distribution of data and equal
powerful, meaning they have a greater likelihood of detecting a true variances among groups.
effect (rejecting a false null hypothesis). • Linear Regression: Examines the relationship between two or more
variables. Assumes normality of residuals, linearity, and
Key Characteristics of Parametric Tests homoscedasticity (equal variance of residuals).
• Data Level: They typically require data measured at the interval or
ratio scale (continuous data).
• Normality: A sample size of 30 or more often makes it acceptable
to assume normality (Central Limit Theorem). For smaller samples,
Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai
 4.2 Parametric and Non-parametric tests Cont’d…

If the data do not meet the assumptions for parametric tests way ANOVA.
(e.g., non-normality, ordinal-level data, or unequal variances), • Spearman’s Rank Correlation: Nonparametric alternative
nonparametric tests provide an alternative. to Pearson’s correlation.
Nonparametric tests are statistical methods used to analyze • Wilcoxon Signed-Rank Test: Nonparametric alternative to
data that do not meet the assumptions required for parametric the paired t-test.
tests. Nonparametric tests do not require the stringent
assumptions of parametric tests. They: The suitable test will be chosen depending on the
• Are suitable for data measured at nominal or ordinal characteristics of the data and study design.
levels.
• Do not assume normality or homogeneity of variance.
• Are less sensitive to outliers or small sample sizes.

Examples of non-Parametric Tests

• Mann-Whitney U Test: Nonparametric alternative to the
independent t-test.
• Kruskal-Wallis Test: Nonparametric alternative to one-
Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai
Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai
 4.3 Independent and dependent t tests

 When the researcher wants to test the significance of difference between

the two means, he/she uses either t-test or z test
 t-test is a parametric test and can be used when the sample size is less
than 30 to test the significant difference between two means.
 William Seely Gosset has developed this test in 1915, he published his
work in the pen name ‘student’ hence it is also called as student t-test.
 The t-test is still valid for larger sample sizes, but for very large samples,
its results closely approximate those of a z-test because of the Central
Limit Theorem.
 There are three main types of t-tests, each designed for different study
designs and data structures:
a. One sample t-test
b. Independent sample t-test
c. Paired sample t-test

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.3 Independent and dependent t tests

1. One-Sample t-Test: Used to compare the mean of a

single sample to a known population mean. Used to see if a
sample differs significantly from a population value. If the
sample size increases then the t distribution gets closer to
the z distributions.
Example: Testing if the average height of male NCC cadets
differs from the national average height of 168 cm.

Formula:

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.3 Independent and dependent t tests

Exercise:

a. A psychologist believes that the average IQ of students in a particular school is different from the general population
mean IQ of 100. A random sample of 15 students from the school is selected, and their IQ scores are as follows:

IQ Scores: 102, 98, 105, 100, 95, 101, 99, 97, 96, 104, 103, 98, 102, 100, 99.

Test if the average IQ of students in this school is significantly different from the population mean (μ=100).

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.3 Independent and dependent t tests

2. Independent sample t test (Two-Sample t-Test): Assumptions:

Used to compare the means of two independent groups • The two groups are independent.
to see if they are significantly different. The two groups • The data within each group are normally distributed.
being compared must be separate and not related. • Homogeneity of variance (similar variances in both
Example: Comparing the test scores of male and female. groups).

Formula:

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.3 Independent and dependent t tests
Exercise:

b. A researcher wants to compare the exam performance of students who studied online versus those who attended in-person
classes. The scores of 10 students in each group are as follows:

Online 78 85 88 89 90 91 74 80 84 80

In-person 82 75 90 78 80 84 83 70 75 95

Determine if there is a significant difference in performance between the two groups.

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.3 Independent and dependent t tests

3. Paired Samples t-Test (Dependent t-Test): Used to Assumptions:

compare the means of the same group or related groups • The differences between paired observations are
under two different conditions. Typically for normally distributed.
measurements taken before and after an intervention
(pre-test and post-test scores of the same participant), or Formula:
when participants are matched in pairs.
Example: Comparing the scores of students before and
after a training program.

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.3 Independent and dependent t tests

Exercise:

c. A fitness trainer wants to evaluate the effectiveness of a 4-week training program. The trainer measures the
participants' weight before and after the program. The weights (in kg) of 8 participants are as follows:

Before 80 85 90 75 100 85 80 86
program

After program 78 81 90 74 95 70 75 84

Test whether the training program significantly reduced participants' weight.

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

4.3 Independent and dependent t tests

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.4 Pearson’s correlation

 Correlational methods are the most commonly used

statistical techniques in the testing fields. Some of the  The direction of the correlation is conveyed by the sign of
common correlational methods are Pearson, spearman, the correlation coefficient and can be a positive correlation
Kendall's tau correlation. (variables change in the same direction), a negative
 Correlation refers to the relationship or association correlation (variables change in opposite directions), or a
between two variables zero correlation (values of one variable are not related to
 The correlation coefficient is a numerical measure that changes in the value of the other variable).
quantifies the strength and direction of a correlation. It is  A correlation only allows us to conclude that two variables
typically denoted as r and ranges from -1 to +1. A covary to some degree. Correlation does not indicate
correlation coefficient of 0 means there is no linear causation.
relationship between the variables.  Scatter plot is the is a graphical representation of the
relationship between two continuous variables. Each point
on the graph represents a pair of values, with one variable
plotted on the x-axis and the other on the y-axis.

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.4 Pearson’s correlation Cont’d…

Types of correlation
1. Positive Correlation: A positive correlation occurs when two variables move in the
same direction. As one variable increases, the other variable also increases, and vice
versa. The correlation coefficient for positive correlation ranges from 0 to +1. A positive
value of the correlation coefficient indicates a direct relationship between the variables.
Example: Generally, as height increases, weight also increases. This is a positive
correlation, as both variables increase together.

2. Negative Correlation: A negative correlation occurs when two variables move in

opposite directions. As one variable increases, the other decreases, and vice versa. The
correlation coefficient for negative correlation ranges from 0 to -1. A negative value of the
correlation coefficient indicates an inverse or opposite relationship between the
variables.
Example: As the amount of exercise increases, body weight may decrease, indicating a
negative correlation.

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.4 Pearson’s correlation Cont’d…
3. Perfect Positive Correlation: Perfect positive correlation means that the two variables have a perfect,
linear relationship. As one variable increases, the other increases in exact proportion, and the relationship
between the variables is exactly predictable. The correlation coefficient for perfect positive correlation is
exactly +1.
Example: Temperature in Celsius and Fahrenheit: The temperature in Celsius and Fahrenheit are perfectly
correlated because the relationship between them is always linear and predictable

4. Perfect Negative Correlation: Perfect negative correlation occurs when two variables have a perfect,
inverse, linear relationship. As one variable increases, the other decreases in exact proportion. The
relationship between the variables is perfectly predictable but in the opposite direction. The correlation
coefficient is exactly -1.
Example: Amount of fuel in a Car and Distance Traveled: As a car travels a fixed distance and consumes
fuel at a constant rate, the amount of fuel in the tank decreases in exact proportion.

5. Zero/No correlation: The two variables are not at all related to each other. As one variable increases
or decreases, it has no relation with the other variable. The correlation coefficient for zero correlation is 0.
Example: Rainfall in America and crop growth in india

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.4 Pearson’s correlation Cont’d…

 Prof. Karl Pearson has given the Pearson’s r correlation method.

 It is the most common methods of assessing the association between two variables under study.
 Also known as Pearson product-moment correlation and abbreviated as r.
 The size of the Pearson r varies from -1 to +1.

Exercise: Calculate product moment coefficient of

correlation between the following marks obtained by the
students in their mathematics and statistics examination

Student Statistics Mathematics

A 4 5
B 8 7
C 9 8
D 5 6
E 2 2

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.4 Regression

One of the important aims of science is to be able to make predict the value of the dependent variable based on known values
predictions. The predictions involve a statement of causality that of the independent variables.
states that X is a cause of Y or X can predict Y. Here, X is the causal Types of regression Based on the Number of Independent
variable or predictor variable or independent variable. Y is the effect Variables:
variable or criterion variable or dependent variable. Often 1. Simple linear regression: It is a statistical method that allows
experimentation is the ideal way to establish a relation of causality us to summarize and study relationships between two
between two variables. However, it is not always possible for the continuous (quantitative) variables:
experimenter to manipulate the IV for practical or ethical. For a. One variable, denoted x, is regarded as the predictor,
example, to study if parent’s height can predict children’s height, one explanatory, or independent variable.
cannot manipulate the height of the parent. Similarly, to study if b. The other variable, denoted y, is regarded as the
levels of marital conflict between a couple can predict depression response, outcome, or dependent variable.
rates, an experimenter, for all ethical purposes, cannot create marital There is only one predictor variable
conflict. Hence, in researches like these, regression can be a way of A simple linear regression uses one IV to predict the DV. For
establishing and studying prediction. example, marks in an exam are predicted from intelligence scores.
Regression analysis is a statistical technique for Assumptions of simple linear regression: Data must be normally
modeling the relationship between a dependent (or response) distributed, there must be a linear relationship between IV and DV,
variable and one or more independent (or predictor) variables. The Independence of observations and homogeneity of variance.
goal of regression analysis is to develop a model that can be used to
Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai
 4.4 Regression

2. Multiple linear regression: It is useful to predict a DV

using multiple predictor variables. For example, marks
obtained in the exam can be predicted from intelligence
quotient, number of hours studied and conscientiousness.
The data must be continuous (quantitative).
Assumptions of simple linear regression:
• Linearity: There is a straight-line relationship between the
dependent and independent variables.
• Normality: The residuals are normally distributed.
• Homoscedasticity: The variance of the errors is the same
across all values of the independent variables.
• No multicollinearity: The independent variables are not too
highly correlated with each other

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.4 Regression

The general form of each type of regression is: the form of a straight line (linear regression) that best
• Linear regression: Y = a + bX + e approximates all the individual data points. In multiple
• Multiple regression: Y = a + b1X1+b2X2 + b3X3 + ... + e regression, the separate variables are differentiated by using
Where: numbers with subscripts.
• Y = the variable that you are trying to predict (dependent
variable).
• X = the variable that you are using to predict Y
(independent variable).
• a = the intercept.
• b = the slope.
• e = the regression residual.

Regression takes a group of random variables,

thought to be predicting Y, and tries to find a mathematical
relationship between them. This relationship is typically in
Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai
 4.5 ANOVA

 ANOVA (Analysis of Variance) is a statistical technique  ANOVA is used to compare the means of multiple
used to determine whether there are significant groups to identify any statistically significant differences
differences between the means of three or more between them. Suppose we want to compare the level of
independent groups. self-esteem in males, females, and transgender
 Prof. R. A. Fisher is the pioneer in developing the individuals. We randomly select samples from each
theoretical concepts of ANOVA. population and use ANOVA to analyze the data.
 Assumptions of ANOVA  Advantages over t-tests: Using ANOVA instead of
1. Independence: Observations are independent multiple t-tests saves time and energy, as it reduces the
of each other number of comparisons needed and it also captures
2. Normality: Data is normally distributed within interaction effects between groups,
each group
3. Homogeneity of variance: Variances are equal
across all groups
4. ANOVA assumes continuous data

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.5 ANOVA

 Purpose of ANOVA: When comparing the means of two  It assumes that sample populations follow a normal
samples, tests such as the z-test or t-test can be used. distribution and have equal variances.
However, when there are more than two sample groups,  The method helps identify whether a particular factor
conducting multiple individual comparisons increases the significantly influences a dependent variable.
risk of errors. ANOVA allows researchers to  The test is performed using the F-ratio, which compares
simultaneously test multiple groups to determine variance between groups and variance within groups. If
whether they originate from populations with the same the calculated F-value exceeds a critical threshold, the
mean. differences between the group means are considered
statistically significant.
Basic principles of ANOVA:
 The fundamental idea of ANOVA is to compare the
means of different populations by analyzing the variation
within each sample relative to the variation between the
samples.

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.5 ANOVA Cont’d…
Types of ANOVA the interaction effect between the two factors.
 One-Way ANOVA: Compares the means of three or  Example: Comparing the self-esteem levels of males and
more groups based on one independent variable. females in different age groups (e.g., 18-25, 26-35). Here
Example: Comparing the self-esteem levels of males, we have two independent variables (gender and age
females, and transgender individuals. Here we have one group) and one dependent variable (self-esteem
Independent variable (at multiple groups/levels) and one
dependent variable.
Independent variable Dependent
variable
Independent variables Dependent
variable
Gender Male
Gender Male
Female Self-esteem
Female
Transgender
Self-esteem
Age group 18-25 years

 Two-Way ANOVA: Compares the means of groups 26-35 years

based on two independent variables. It also considers

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.5 MANOVA Cont’d…

 The Multivariate Analysis of Variance (MANOVA) is an Why Use MANOVA?

extension of ANOVA that allows researchers to
 Accounts for multiple DVs simultaneously instead of
compare multiple dependent variables (DVs) at the
conducting separate ANOVAs, which reduces the risk
same time. While ANOVA examines differences in one
of Type I errors.
dependent variable across groups, MANOVA is used
when there are multiple interrelated dependent
 Considers relationships between dependent variables
(e.g., anxiety and fatigue are often correlated).
variables that need to be analyzed simultaneously.

 For example, in psychological research, MANOVA can  More powerful than multiple one-way ANOVAs in
detecting differences between groups.
be used to study whether different groups show
significant differences in depression, anxiety, and stress
levels at the same time.

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

 4.5 MANOVA Cont’d…

Types of MANOVAs Assumptions of MANOVA

 One-Way MANOVA: Used when there is one independent  Normality: Each dependent variable should be normally
variable (IV) with multiple levels affecting multiple distributed.
dependent variables. Example: Studying coffee consumption
 No Univariate or Multivariate Outliers: Extreme values
(IV) and its effects on anxiety and fatigue (DVs).
should be checked and handled.
 Two-Way MANOVA: Used when there are two
 Multicollinearity: The dependent variables should be
independent variables affecting multiple dependent
correlated but not too strongly (to ensure they measure
variables. Example: Studying coffee consumption and sleep
distinct concepts).
quality (IVs) and their effects on anxiety and fatigue (DVs).
 Linearity: There should be a linear relationship between
 Mixed MANOVA: Used when a study involves both
dependent variables.
between-subject and within-subject factors. Example:
Measuring anxiety and fatigue in participants before and  Homogeneity of Variance-Covariance Matrices: The variance
after drinking coffee (within-subjects) while comparing structure across groups should be similar.

between coffee and water groups (between-subjects).

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai
 4.5 MANOVA Cont’d…

MANOVA is a powerful statistical method for analyzing multiple dependent variables

simultaneously. It helps psychologists reduce errors while identifying meaningful patterns in data.

It is particularly useful when dependent variables are related and allows researchers to gain

deeper insights into complex human behaviors

Dr. Amani Kanugolu, Assistant professor, NMIMS, Navi Mumbai

THANKS !