EXTRA

Question 1

When do we use a quasi-experiment?

Answer 1

When we cannot run a true experiment. If you do not have permission to randomly assign people to groups.

So, when we cannot run a field or lab experiment.

Question 2

If you would study Netflix the same way as Datta, Knox, and Bronnenberg (2018), how would you set up the study?

Answer 2

1. Find consumers who have and have not adopted Netflix.
2. Match the adopters with the non-adopters based on likelihood to adopt.
3. Investigate how many (different types of) movies they consume.
4. Investigate this with a difference-in-differences analysis to see the change in consumption due to the Netflix adoption.

Question 3

Standard error of the coefficient

Answer 3

Measures how precisely the model estimates the coefficient’s unknown value.

The smaller the standard error, the more precise the estimate.

Question 4

What can be concluded from a positive correlation?

Answer 4

That as the independent variable increases, the dependent variable also tends to increase.

Question 5

p-value > significance level

p-value is higher than the significance level

Answer 5

It is not statistically significant. The influence of this independent variable could have been due to random chance.

There is insufficient evidence to conclude a relationship. We cannot reject the null hypothesis.

Question 6

p-value < significance level

p-value is lower than the significance level

Answer 6

It is statistically significant.

There is enough/sufficient evidence that there is a relationship between the IV and the DV. We reject the null hypothesis.

Question 7

What do the p-values indicate?

Answer 7

Whether the relationships are statistically significant.

If the relationship you observe in the sample also exists in the larger population.

Question 8

What do coefficients describe?

Answer 8

The mathematical relationship between each independent variable (IV) and the dependent variable (DV).

Question 9

Long-tail concept

Answer 9

A business strategy that allows companies to realise significant profits by selling low volumes of hard-to-find items to customers, instead of only selling low volumes of a reduced number of popular items.

Question 10

Two factors that indirectly influence sales:

Answer 10

Consumer-brand usage (CBU)

Consumer-brand endorsement (CBE)

Question 11

Propensity Score Matching

Answer 11

Quasi-experimental method in which the researcher uses techniques to construct an artificial control group by matching each treated unit with a non-treated unit of similar characteristics.

Question 12

Difference supervised and unsupervised machine learning.

Answer 12

Supervised learning uses labeled input and output, while an unsupervised learning algorithm does not.

Question 13

Unsupervised learning method

Answer 13

Machine learning algorithm that analyses and clusters unlabelled datasets.

These algorithms discover hidden patterns in data without the need for human intervention. Hence, they are unsupervised.

Question 14

Supervised learning method

Answer 14

Machine learning approach that’s defined by its use of labeled datasets.

Using labelend inputs and outputs, the model can measure its accuracy and learn over time.

Question 15

Machine learning methods can be subdivided into:

Answer 15

• Supervised methods

- Unsupervised methods

Question 16

Gross Rating Point (GRP)

Answer 16

A ratio that measures the influence of advertising campaigns on their target audience.

Question 17

Shapley Values Method

Answer 17

Method with which you are looking at comparable ‘paths to purchase’, with the difference being in that one channel is missing.

Question 18

Limitations of PROSAD with respect to finding the impact a bid has on profitability.

Answer 18

• It does not provide information on what would have happened if no bid was placed.
• It does not take organic search results into account.
• It does not (directly) include competition.

Question 19

What is PROSAD

Answer 19

A tool that investigates the effectiveness of Search Engine Advertising.

Question 20

Problems occurring when measuring ROI based on clicks, sales, or visitor numbers instantly generated by a campaign.

Answer 20

1. Measuring the month-to-month impact of digital marketing is far too short a timeframe to assess the true impact.
2. Basic measures such as visitor numbers, click-through rate, CPC, or CPL don’t tell business leaders anything about the actual impact of a campaign since they are slow-burn projects, starting with customer engagement and only months later generating sales.

They are useful, but rather seen as KPIs, and not as a measure of ROI.

Question 21

What can be concluded from a positive correlation?

Answer 21

That as the independent variable increases, the dependent variable also tends to increase.

Question 22

What can be concluded from a negative correlation?

Answer 22

That as the independent variable increases, the dependent variable decreases.

Question 23

Interpreting r-squared

Answer 23

The larger the r-squared, the better the model explains the variations.

Higher r-squared: the more the dotted lines are closer to the regression model line.

Lower r-squared: the further away the dotted lines are from the regression model line.

Question 24

What does the adjusted r-squared determine?

Answer 24

It determines how reliable the correlation is and how much it is determined by the addition of independent variables.

Question 25

What tends to increase the r-squared value?

Answer 25

Adding more independent variables.

But this can also lead to overfitting.

Question 26

Meaning coefficient value

Answer 26

Indicates the mean change in the dependent variable given a one-unit change in the independent variable, while holding the other variables in the model stable.

Question 27

What can be concluded from a positive correlation?

Answer 27

That as the independent variable increases, the dependent variable also tends to increase.