29. august 2012 - lecture #2 (Variables and operationalization of them)

Task for next time: Deliver an essay by the 5th of September in Fronter. Task is described on page 8 of the lecture notes. (Describe variables (independent, dependent and operationalized), validity and reliability, in the "Kansas is flatter than a pancake" and "Portable devices as visual noise during lectures" papers

Sources for research ideas:

• Daily life curiosity
  
• Generally faults or when things go wrong: why?
  
• Past research and theorybased
  

Process:

• Formulate the research problem. A question!. Avoid what you think of as the solution as a question, and limit the scope to something you have resources to complete.
  
• Define variables involved:
  Independent variables are the variables(s) of the situation you try to control/manipulate in order to conclude either a correlation or causation with other variables. The manipulation can be done by the experimenter or it can be a "natural experiment" like a natural phenomena. The assumed cause.

  Dependent variables are all other variables, and the independent variable might directly or indirectly cause changes in them. The assumed effect.

  The text book talkes about mediating and moderating variables. Mediation variables are "in between", and can be thought of as actual reasons for why something happens and is often influenced by the independent variable. Moderating variables can be totally independent of the independent variable and cause amplification (positive or negative) of the outcome (dependent variables tested for). Like context/conditions.
  

• Operationalization (operational definition): How the variables (if not obvious) will be measured.
  

Example:
Research problem: "Did Lillehammer-OL make people in this region more happy”?
Define “Lillehammer-OL” as independent variable. Dependent variable would be happy. You then figure how to compare the happy measure with/without (before/after) the stimuli of "Lillehammer-OL". The variable happy must be defined in a precise way: One example from the original researchers of this problem were: “Count the number of smiles in the main shopping street”

Research problem: "Does attitude campaign reduce speeding on a given road?"
Independent variable is "campaign" (or "no campaign"): Dependent variable could be number of speeders (with speed above speed limit in a given time period).

Research problem: "Does meditation make better humans?"
Independent variable is "meditation" (or "no meditation") What kind of meditation? How much meditation?. Dependent variable is "better human". What is a better human?

Hypothesis
The expected results, the best guess or prediction of what will happen. Define criteria of H1 or H0

Causation vs correlation

• A correlation is simply when two variables are affected by the other. Variable A increases or decreases the value of variable B in a predictable way.
  
• Determining causation is much harder. It requires control of all correlated variables. You need to have the one independent variable be the only variable with change (except for the dependent variable you are interested in), and you still might measure the cause indirectly. One way to do this is with a "control group", someone or something that does not get the influence of the independent variable.
  

An independent variable X, The observed dependent variable Y an unknown variable Z: Their interactions

..not quite sure about the underlying though...

Threats to causation:

• Changes because of time (with humans maturation)
  
• Changes in measurement equipment.
  
• Direction: A → B or B → A?
  
• Unknown C → (A and B)
  
• Unknown C → A → B
  
• A is a co-factor or facilitator (A and unknown C → B)
  

Confunding variables are the variables we don't have control over or even don't know about. Like "Z".

“real world is messy” → isolate in lab as a first step (You might not capture all relevant variables). Record as much as possible

Reliability and validity

• Reliability: A measure of how consistent results are (how precise you are) (Cronbach's alpha). Retesting, testing with different subjects are ways to increase reliability.
  
• Validity: A measure of how close the results are to the actual world, on target (what you do). Internal validity (in this case) or external validity (globally true)