Introduction to Research in Marketing 2018-2019
Table of Contents
Introduction (Lecture) ................................................................................................................................... 2
Introduction (Video Lecture) ........................................................................................................................ 4
ANOVA (Lecture) ........................................................................................................................................... 6
ANOVA (Video) .............................................................................................................................................. 9
Video 1 ...................................................................................................................................................... 9
Video 2 .................................................................................................................................................... 12
Cluster Analysis (Lecture)............................................................................................................................ 13
Cluster Analysis (Video Lecture) ................................................................................................................. 16
Factor Analysis (Lecture) ............................................................................................................................. 20
Factor Analysis (Video)................................................................................................................................ 23
Logistic Regression (Lecture) ...................................................................................................................... 26
Logistic Regression (Video Lecture) ............................................................................................................ 28
Conjoint Analysis (Lecture) ........................................................................................................................ 31
Conjoint Analysis (Video Lecture) ............................................................................................................... 34
Multidimensional Scaling (MDS) (Lecture) ................................................................................................. 42
Multidimensional Scaling (MDS) (Video) .................................................................................................... 47
Similarity-based approach ...................................................................................................................... 47
Attribute-based approach....................................................................................................................... 49
Summarizing Table ...................................................................................................................................... 52
Practice Exam .............................................................................................................................................. 54
Practice Exam Answers ............................................................................................................................... 56
Theory Practice Exam.................................................................................................................................. 59
Answers ....................................................................................................................................................... 62
Introduction to Research in Marketing Fall 2018-2019 Eline van de Ven
,Introduction (Lecture)
Who I ask How I analyze it
The total error framework
What you observe = True value + Sampling error + Measurement error + Statistical error
How I ask it
If you mess up any of these errors, your results will be biased.
A sample is a representative of a population.
Population
Sample
Statistics Parameters
Characteristics Characteristics of population
of sample (e.g. mean 𝜇)
(e.g. average 𝑥̅ )
Target population Frame population Sample population Respondents
Coverage error Sampling error Non-response error
𝑁1 N2
Weighted mean = 𝑥̅ 1 + 𝑥̅ 2
𝑁1+𝑁2 N1+N2
N1 = biased sample
N2 = looks more like population
Non-Metric
1) Nominal (Do I buy A or B?)
2) Ordinal (Do I like it a lot or a little?)
Metric
1) Interval
2) Ratio (most precise!)
Validity: Does it measure what it’s supposed to do? Face validity: Do these coefficients make sense?
Reliability: Is it stable? How much do these results change if
• We add additional control variables to the model
• We take away some observations
• We estimate the same model on a new dataset
Introduction to Research in Marketing Fall 2018-2019 Eline van de Ven
, Type 2 error: you say H0 but it Alt is true
Power
Type 1 error: you say it is Alt but null is true
1 – P (Null | Alt) = Power
Type 1 and 2 errors are inverse related. When type 1 error decreases, type 2 will probably increase.
Lowering the type 1 error decreases the power.
The bigger the size, the bigger the power. Power is how big the chance is that you find Alt.
When your study is underpowered, increasing the sample size will exaggerate the effect.
Explore data;
• recode missing (9999 is missing)
• Reverse code negatively worded questions
• Correct range
• Consistency Age 22? Then birthyear 1996?
Outliers are observations with a unique combination of properties that differentiate them from other
observations.
Introduction to Research in Marketing Fall 2018-2019 Eline van de Ven
,Introduction (Video Lecture)
How to split variables:
1) Transform, compute variable
2) Call the new variable “Date_String” Set the width to 8 (01/03/96 is 8 signs)
3) Function = String, Type = String
After that we have to find out how to split the variables right, since the date and time don’t always
begin on the same spot. You can do this by entering this into the compute variable part:
The 1st sign until the last one, in this
case where the space is (which is not
always the same, therefore we use
4) CHAR.SUBSTRACT.(3) another CHAR formula
(TIMESTAMP,1,CHAR.INDEX(2) The Space
(Timestamp, ‘ ‘)
So the complete formula is:
CHAR.SUBSTRACT(TIMESTAMP,1,CHAR.INDEX(TIMESTAMP,’ ‘)
Next get SPSS to recognize it as a “date” variable instead of a string variable. (Transform, Date and Time
Wizard, Create a date/time from a string, “mm/dd/yy”). Call this “date”.
When you want to split the rest of the date&time variable, you need to know at what number of signs
the time part starts.
You can do this by changing the previous formula to:
CHAR.SUBSTR(Timestamp,CHAR.INDEX(Timestamp,' ')+1)
How can we find out what hour of the day most calls are made in Amsterdam? (Use dataset assignment
1)
1) Split cases – Transform – recode into different
2) Select cases if AMS = 1, and delete unselected
3) Analyze, descriptive, frequencies “Hour”
Introduction to Research in Marketing Fall 2018-2019 Eline van de Ven
, How can I merge files?
1. Let the files ascend in the same way.
2. In the Variable view, make sure the variable of interest has the same width in both files
3. Merge so that for each city in the response file, we see its unemployment and retirement
statistics. (Data, merge, add variables, open dataset, match cases on key variables, move city
from excluded to key variables, cases are sorted , non-active dataset is keyed table.)
Missing values
How many observations are missing? Analyze, missing value analysis
Missing data occurs when values of one or more variables aren’t found. This often happens with
multivariate analyses. Missing data results in 1) reduction of the sample size that is available to analysis,
and 2) bias when statistical results are based on the missing data. When missing data is 15%, it can be
deleted. Higher levels (20%/30%) have to be ‘helped’.
Outliers
Observations with a unique combination of characteristics that differentiate themselves from other
observations. Usually this is an extremely high or low value or a unique combination of values. It can
also occur due to a procedural error (e.g. data error, wrong coding). The outlier has to be looked at in
terms of how representative it is to the population.
Introduction to Research in Marketing Fall 2018-2019 Eline van de Ven
Table of Contents
Introduction (Lecture) ................................................................................................................................... 2
Introduction (Video Lecture) ........................................................................................................................ 4
ANOVA (Lecture) ........................................................................................................................................... 6
ANOVA (Video) .............................................................................................................................................. 9
Video 1 ...................................................................................................................................................... 9
Video 2 .................................................................................................................................................... 12
Cluster Analysis (Lecture)............................................................................................................................ 13
Cluster Analysis (Video Lecture) ................................................................................................................. 16
Factor Analysis (Lecture) ............................................................................................................................. 20
Factor Analysis (Video)................................................................................................................................ 23
Logistic Regression (Lecture) ...................................................................................................................... 26
Logistic Regression (Video Lecture) ............................................................................................................ 28
Conjoint Analysis (Lecture) ........................................................................................................................ 31
Conjoint Analysis (Video Lecture) ............................................................................................................... 34
Multidimensional Scaling (MDS) (Lecture) ................................................................................................. 42
Multidimensional Scaling (MDS) (Video) .................................................................................................... 47
Similarity-based approach ...................................................................................................................... 47
Attribute-based approach....................................................................................................................... 49
Summarizing Table ...................................................................................................................................... 52
Practice Exam .............................................................................................................................................. 54
Practice Exam Answers ............................................................................................................................... 56
Theory Practice Exam.................................................................................................................................. 59
Answers ....................................................................................................................................................... 62
Introduction to Research in Marketing Fall 2018-2019 Eline van de Ven
,Introduction (Lecture)
Who I ask How I analyze it
The total error framework
What you observe = True value + Sampling error + Measurement error + Statistical error
How I ask it
If you mess up any of these errors, your results will be biased.
A sample is a representative of a population.
Population
Sample
Statistics Parameters
Characteristics Characteristics of population
of sample (e.g. mean 𝜇)
(e.g. average 𝑥̅ )
Target population Frame population Sample population Respondents
Coverage error Sampling error Non-response error
𝑁1 N2
Weighted mean = 𝑥̅ 1 + 𝑥̅ 2
𝑁1+𝑁2 N1+N2
N1 = biased sample
N2 = looks more like population
Non-Metric
1) Nominal (Do I buy A or B?)
2) Ordinal (Do I like it a lot or a little?)
Metric
1) Interval
2) Ratio (most precise!)
Validity: Does it measure what it’s supposed to do? Face validity: Do these coefficients make sense?
Reliability: Is it stable? How much do these results change if
• We add additional control variables to the model
• We take away some observations
• We estimate the same model on a new dataset
Introduction to Research in Marketing Fall 2018-2019 Eline van de Ven
, Type 2 error: you say H0 but it Alt is true
Power
Type 1 error: you say it is Alt but null is true
1 – P (Null | Alt) = Power
Type 1 and 2 errors are inverse related. When type 1 error decreases, type 2 will probably increase.
Lowering the type 1 error decreases the power.
The bigger the size, the bigger the power. Power is how big the chance is that you find Alt.
When your study is underpowered, increasing the sample size will exaggerate the effect.
Explore data;
• recode missing (9999 is missing)
• Reverse code negatively worded questions
• Correct range
• Consistency Age 22? Then birthyear 1996?
Outliers are observations with a unique combination of properties that differentiate them from other
observations.
Introduction to Research in Marketing Fall 2018-2019 Eline van de Ven
,Introduction (Video Lecture)
How to split variables:
1) Transform, compute variable
2) Call the new variable “Date_String” Set the width to 8 (01/03/96 is 8 signs)
3) Function = String, Type = String
After that we have to find out how to split the variables right, since the date and time don’t always
begin on the same spot. You can do this by entering this into the compute variable part:
The 1st sign until the last one, in this
case where the space is (which is not
always the same, therefore we use
4) CHAR.SUBSTRACT.(3) another CHAR formula
(TIMESTAMP,1,CHAR.INDEX(2) The Space
(Timestamp, ‘ ‘)
So the complete formula is:
CHAR.SUBSTRACT(TIMESTAMP,1,CHAR.INDEX(TIMESTAMP,’ ‘)
Next get SPSS to recognize it as a “date” variable instead of a string variable. (Transform, Date and Time
Wizard, Create a date/time from a string, “mm/dd/yy”). Call this “date”.
When you want to split the rest of the date&time variable, you need to know at what number of signs
the time part starts.
You can do this by changing the previous formula to:
CHAR.SUBSTR(Timestamp,CHAR.INDEX(Timestamp,' ')+1)
How can we find out what hour of the day most calls are made in Amsterdam? (Use dataset assignment
1)
1) Split cases – Transform – recode into different
2) Select cases if AMS = 1, and delete unselected
3) Analyze, descriptive, frequencies “Hour”
Introduction to Research in Marketing Fall 2018-2019 Eline van de Ven
, How can I merge files?
1. Let the files ascend in the same way.
2. In the Variable view, make sure the variable of interest has the same width in both files
3. Merge so that for each city in the response file, we see its unemployment and retirement
statistics. (Data, merge, add variables, open dataset, match cases on key variables, move city
from excluded to key variables, cases are sorted , non-active dataset is keyed table.)
Missing values
How many observations are missing? Analyze, missing value analysis
Missing data occurs when values of one or more variables aren’t found. This often happens with
multivariate analyses. Missing data results in 1) reduction of the sample size that is available to analysis,
and 2) bias when statistical results are based on the missing data. When missing data is 15%, it can be
deleted. Higher levels (20%/30%) have to be ‘helped’.
Outliers
Observations with a unique combination of characteristics that differentiate themselves from other
observations. Usually this is an extremely high or low value or a unique combination of values. It can
also occur due to a procedural error (e.g. data error, wrong coding). The outlier has to be looked at in
terms of how representative it is to the population.
Introduction to Research in Marketing Fall 2018-2019 Eline van de Ven
