session 4 Flashcards
by source
primary and secondary data
by methodology
qualitative and quantitive
by objective
descriptive, exploratory, causal/ experiment
Exploratory research:
seek to define an ambiguous problem
- May be conducted as part of the problem definition
- Flexible and adaptive
- If you wanted to understand the root problem for j crew
- usually use quantitative methods
- use primary data or secondary data
primary vs secondary data
- primary data: data that is directly from the source (data you collect, usually done in a qualitative way for exploratory research)
- Secondary data you get from another source (go onto a website and you use stats from third party websites that is owned by another party)
Exploration is complete when
The problem is fully defined
Root of problem, not just symptoms
No more whys
Can state research problems and objectives
No additional information is available
Its a judgment call
Can always engage in more expiration research
If you are looking at factors
Descriptive research
Seek to describe the target market
Answers to questions of who what where when and how
More rigid than exploratory research → can use more quantitative methods to examine consumers
Example: better understand average consumer
Options to conduct descriptive research:
observation and survey
observation and survey
Observation: watch how people interact with environment in rea life situations or the lab
At mcgill we have a couch tarde store that is used for research they put eye trackers on people and can see the whole customer journey and what catches their eye, their fixation on price, etc
Survey: ask people about attitudes, beliefs, behaviours, reactions
Can use surveys to describe consumers and find out as much info about them as possible
Causal research
Seek to answer a defined and described problem
Determine causality
Most rigid → not bad just a rigid way
whether one variable causes another → determine through an experiment
Example: does decreasing sugar content affect sales → want to determine the causal relationship between the two
You are trying to determine causality and if sugar content effects sales
Option: experiment
Examine differences between control; & experimental groups under controlled conditions
causal methods
experiment = making causal claims
If i change pricing before christmas and my sales spike can I infer that my pricing changes were the cause?–> cannot infer that unless you are in an experiment where you isolate price
In response to the introduction of a new product by my competitor, i re-positioned my product. After that i notice a decrease in sales, can i infer my new positioning is bad? → need to isolate the effect of my positioning, need to isolate positioning on sales
need an experiment and need to rule out confounding effects
confounding effects:
when you have another variable that can explain your effect. There can be many reasons why your sales spike it might not just be your price (other variables may explain the relationship between 2 variables, this is confounding effects)
correlation vs causation
correlation= two variables share some kind of relationship
causation= one variable that causes something to happen in another variable
Correlation can be explained by:
- One-way causality: one variable is the cause of the other one (when X causes Y)
- Two way causality: both variable may be the cause of each other (when X causes Y and Y could cause X)
- A confound: or third variable may be responsible for the correlation. X is associated with y and can be explained by the confound (Z variable is the confound and affects both X and Y variables)
- Spurious correlation: a mathematical relationship in which two events or variables have no causal relationship ( x and Y correlate randomly, doesnt mean its meaningful)
Number of people who drowned into a falling pool and films nicholas cage appeared in=
spurious
Minutes spent studying and score on test= positive relationship
Would be one way causality and confound
would be one way causality and cant be two way because you cant go back in time but if you do well on test 1, you might change your studying but this example just looks at one test
Confound: if it’s an experiment need more info (one way causality is found), if its simply an association and no experiment then yes we can determine there may be a confound (may be certain psychological characteristics at play that affect you test scores)
Eating chocolate and nobel laureates produced
Theres a positive relationship
Spurious correlation and confounds
There could be multiple variables that explain the relationship, more or less a spurious correlation because there is no logical reason for this or scientific basis
Watching too much tv is bad for health
Could be two causality and confounds
Experimental design variables
Dependent variable= the effect (for example attitude)
Manipulated variable (independent variable) = the cause (for example product positioning)
Experiment
You have a hypothesis that x has a causal impact on y
Independent variable: antecedent event or cause (x1): a variable sysmetialy vared by the experimenter (fo example: change in price, change in messaging strategy )
Dependent variable: posterior event or effect (Y)
The variable that the experimenter measures (ex: response: consumer attitudes, buying intentions)
Y= X1+X2+X3….+Xt → all the Xs are potential confounds
Level 1: what the product is currently priced at
Level 2: the new price you want to implement and see if it impacts sales
Levels are the different conditions you have for your independent variable
Have to randomly assign participants to conditions a and b
Due to randomization we can look at causal impacts of the independent variable on the dependent variable
examples
Does having a college education impact starting salary?
Correlation, (cant control if people have college education or not) unethical to prevent some from education
Does working shorter hours increase productivity?
Experimental because you can assign people to work different hours and offices and how this affects productivity
Does using comic sans on a resume decrease likelihood of hiring
Can randomly assign people so its experimental can actually test this
- Does being left handed increase creativity?
Correlational cant assign or manipulate if someone is left or right handed
Internal validity:
are the findings due to the independent variable
Sometimes there is no true randomization ( ex: some field experiments)
Need true randomization and large sample sizes
external validity
can I generalize the results to another group or another context
experiments with convient samples
weird samples
need to replicate findings with different populations and different contexts
field exoeirnents and online experiments can offer solutions
ecological valdiity
does my study mimic what would happen in real life
issue with hypothetical studies
field exoeirnents and online experiments can offer solutions