MKMN 1018

Data Analytics and Data Mining

Schedule for Today

Agenda
• t test: 3 types
• ANOVA
• Linear Regression
• Test Statistic review
WEEK 5
T TEST
General Hypothesis Testing Steps

A hypothesis is a premise or a claim that we want to test.

Formulate the Choose the Select the

significance Collect Data Calculate the
Hypothesis appropriate
level 𝞪 test statistic
test*

Compare p
Determine the value
P value to chosen
significance
level 𝞪

*t value, z value, f value, x2 value

 The type of test statistic you use depends on a couple of factors
We will be using excel to perform all calculations.

Type of Data: Categorical Data Numerical Data

Comparison Association Comparison Association

Business Objective:
Comparing Determining Comparing Determining
independent groups relationships independent groups relationships

Test Statistic: Proportion z test Chi-Squared (𝛘2) t test: Independent, one- Linear regression
for binary - Goodness of sample or paired
categorical data fit (1 variable)
- Test of z test: one sample or two
independence sample
(2 variables)
ANOVA (2+ samples)

Key Assumption: Your data is normally distributed (check by plotting in a histogram and observing
the shape)
 #4: What if our sample size is small (<30)…use a t test
Types of t tests:
When to use this test:
When comparing means of a small sample (n < 30)

Test Assumptions:
Sample size is small (>30)
Data is normally distributed
Population variance is unknown – so sample variance is
used
 #5: What if we have a more than 2 variables?
Analysis of Variance (ANOVA): Single Factor When to use this test:
When comparing means of a 3+ small samples (n >
Use sample data to make inferences about the 30).Only looking at single factor for this course
properties of entire populations.
Test Assumptions:
H0:μ1= μ2 = μ3 Sample size is small (>30)
HA: At least one μ differs from the others Data is normally distributed
Population variance is unknown – so sample variance is
used

Example:
You are testing 3 brands of batteries to see what the
average life of a battery is.

You use 18 flashlights and test the 3 brands.

Your goal is to determine the level of variation

between group and within groups.
 #5: What if we have more than 2 variables?
Analysis of Variance (ANOVA): Single Factor

Test Steps:
1. Determine your Null and Alternative Hypothesis
2. Gather your data and ensure you meet all assumptions
3. Use the Data Analysis tool in excel to output the Anova: Single Factor Analysis (F test)
4. Compare the p value to your alpha value
 The type of test statistic you use depends on a couple of factors
We will be using excel to perform all calculations.

Type of Data: Categorical Data Numerical Data

Comparison Association Comparison Association

Business Objective:
Comparing Determining Comparing Determining
independent groups relationships independent groups relationships

Test Statistic: Proportion z test Chi-Squared (𝛘2) t test: Independent, one- Linear regression
for binary - Goodness of sample or paired
categorical data fit (1 variable)
- Test of z test: one sample or two
independence sample
(2 variables)
ANOVA (2+ samples)

Key Assumption: Your data is normally distributed (check by plotting in a histogram and observing
the shape)
 #6: What if we want to determine the relationship between
numerical variables?
Linear Regression
Assumptions
Models the relationship between 2 variables by
1. For any fixed value of x, y is normally distributed
fitting a linear equation to the observed data. 2. Observations are independent of each other
3. The proposed relationship between x and y is linear
y = mx + b

Y is the dependent variable

X is the independent variable Test Steps:
m is the slope of the line
1. Scatterplot to visually determine if a relationship exists
b is the y intercept (value of y when x = 0)
2. Create a Null and Alternative Hypothesis. Null
hypothesis is that there is no correlation.
3. Typically a two tailed test (the relationship being
positive or negative does not usually matter)
4. Use the data analysis tool to determine the p value
5. Check the p value against your alpha value
 Putting it all together….

Match the scenario to the best test statistic and explain your choice:

1. A marketer is looking at customer satisfaction before and after the launch of a new online customer
service portal. 20 people were asked to rate their satisfaction at two different points in time.
2. A sales team is looking at the impact that three variables have on the average transaction size
3. A customer service team is trying to understand if there is any link between customer satisfaction
(Satisfied Yes/No) and gender (Male/Female).
4. A marketing team is trying to determine if there is a relationship between average transaction size
and the purchase time of day

Chi Squared Test

Paired t test Linear Regression ANOVA
for Independence
Week 5 Recap
What we did this week:
1. Test Statistics
- t test
- ANOVA
- Linear regression

Looking ahead to next week:

2. Deep dive on surveys and coding of data
3. Getting ready for Test #1

Action items:
4. Catch up on chapter readings and review concepts from class
5. Review test statistics and bring any questions with you to class next week.