What you need to know to pass the exam.
Types of Research
Types of Research Methods
- Observational - observe and record
- Correlational - relationship b/w 2 variables
- Experimental - consuct a lab experiment
A question about an area of interest. Eg. Is the proportion of women in design higher than in engineering?
A statement about the relationship between 2 variables. Eg. There is no difference between proportion of women in engineering and design.
An experiment is conducted to find the effect of one variable on another.
- The number of hours of sleep the night before an exam, and the marks scored.
- Two different designs of a chair, and the average rating on a scale of 1 to 10 when people were asked to rank comfort.
Independent variable (or Manipulated variable)
The variable we can control. Eg. hours of sleep, design of chair.
Dependent Variable (or Response variable)
The variable that varies because we changed the independent variable. Eg. marks in test, comfort rating
Circumstance not under investigation that is kept constant while testing.
Circumstance not under investigation that is allowed to vary randomly.
Circumstance not under investigation that has impact on the dependent variable.
So we can say that experiment is conducted to find the effect of changing independent variable on dependent variable, but we might get a fake result because of the confounding variable also impacts the dependent variable.
Levels of a variable
The number of possible states of a variable are called the levels of the variable.
For eg. in our experiment we can have 3, 5, 7, 9 hours of sleep, so hours of sleep variable will have 4 levels.
- A is given design 1. They rank it 5/10.
- B is given design 2. They rank it 6/10.
We can say that B is better, but it’s possible that B was just more generous with the rating and would have given design 1 a 6 as well.
- Subject A is given design 1. They rank it 5/10.
- Subject A is given design 2. They rank it 6/10.
Order effects within subjects
Design 1 and 2 share some elements. If you learn to use 1, 2 is easier to use. Therefore if you use 1 first or 2 first will impact the ranking.
Therefore, we give half of the people design 1 and
Goodness of fit Tests
How well does this distribution match an expected distribution?
Z test (Standard error of proportions test)
For binomial variables only