Lecture 2 Flashcards
1
Q
Inferential vs. Descriptive stats
A
- Inferential: Inferences between two different groups
- Understanding the impression/noise between groups
- Descriptive: Just about this particular group
2
Q
What can cause noise?
(and how can we reduce it)
A
- Measurement error
- Population used
- Low sample size
- Reduced by increasing N, and using random assignment.
3
Q
4 Types of Scores/Scales
(and examples)
A
- Nominal: A name, no value.
- Group A, Group B
- Can create frequency distributions
- Ordinal: Has magnitude but no equal intervals or absolute 0
- Ranking from highest to lowest (tallest to shortest rank)
- Likert scale is ordinal
- Doesn’t say anything about distance between them
- Can be manipulated using arithmetic
- Interval: Equal intervals but no absolute 0
- Most psychological tests are interval
- ex. Celsius and Farenheight
- Can apply any arithmetic operation to the differences between scores
- Ratio: Has a true 0
- Calvin
4
Q
Define:
- Inferences
- Measurement
- Covariance
- Dichotomous (Artificial/True) variables
A
- Inferences: Logical deductions about events that cannot be observed directly.
- Covariance: How two measures covary or vary together
- Dichotomous (artificial): made up variables, (true): actual objective differences in variables (e.x. right/wrong)
5
Q
3 Properties of a Scale
A
- Magnitude: Attribute that must be able to be more or less
- Team 1 and 2 do not have a magnitude.
- Equal Intervals: Difference between two points
- Most tests do not have equal intervals (e.x.IQ)
- When it does have equal intervals, it can be described with linear equations
- Absolute 0
- Difficult/impossible to achieve in psychology
6
Q
Percentile Rank
A
7
Q
Variance
And Standard Deviation
A
- Standard Deviation = sqrt of Variance
- Measure variation is similar to finding average deviation around the mean
- Variance is the avg squared deviation around the mean
8
Q
Standard Deviation
A
9
Q
Sample Standard Deviation
A
10
Q
Z-Score
A
- Find th emean of the raw scores
- Find the standard deviation
- Change the raw scroes to z-scores
- Z-scores can now be converted into any other metric
11
Q
Normal Distribution Graph
(and conversions)
A
12
Q
McCall’s T
A
13
Q
Quartiles and Deciles
A
- 1st Quartile = 25%
- 2nd Quartile = 50% median
- 3rd Quartile = 75%
- Deciles = 10% increments
14
Q
Norm-referenced test (and problems) vs. Criterion-Referenced test
(and Age-related norms)
A
- The performances by defined groups on particular tests
- Compare a score to some other distrubtion (standardized)
- Problem: They changed over time and you cant test everyone. Determining what the right group of people is is also difficult
- Age-related: used in IQ tests (compare mental age to actual age)
- Criterion: describes the specific types of skills/task/knowledge that a test taker can demonstrate (e.x. math skills)
- Results of test would not be used to make comparison between others.
- Emphasizes the diagnostic use of tests to identify problems
15
Q
Box: Within-Group Norming Controvery
A
- Different racial and ethnic groups do not have the same average level of perfromance on tests
- Overselection: selecting a higher percentage from a particular group than would be expected.
- Separating norms for different ethnic groups elminates the problem but created new practical ones (and it became illegal to do this)
16
Q
Tracking
A
- The tendency to stay at about the same level relative to one’s peers.
- E.x. 3rd percentil height infant tend to be around the 3rd percentile later in life
- Worked well in medicine, but controversial in education
17
Q
Box: Within High-School Norms for University Admission
A
- Admissions did not reflect demographic characteristics of the state
- Developed a program which guarantees eligibility for top 4% of high school graduates and not on SAT test
- Latino acceptance rates dropped from 68-45 and African from 58-35%
18
Q
Box: No Child Left Behind and (STAR) system in California
A
- Congress passed legistlation to require greater accountability for school performance (made info about school distracts public.
- Each child was required to test proficiency in reading and math in grade 3-8.
- Problems: Tests effect school funding, passing is defined by arbitrary cut point and caused teachers to “teach the test” not the concept.
STAR system
- Evaluates performance for programs like no child left behind
- Not many 3rd graders are advanced, but many more 4th graders are advanced
- Explanation: Definition of advanced is arbitrary. A test too hard for 3rd graders may be too easy for 4th graders.
19
Q
Correlation Coefficient
(Pearson Product Moment Correlation)
A
- Correlation is regression with the scores normalized around -1 and 1 with no intercept
20
Q
Regression
(What does it try to do and why?)
A
- Try to predict one variable from another
- Why? Because x may be easier to calculate than y
- Found by using principle of least squares
21
Q
Correlation Types
A
22
Q
Principle of Least Squares
A
- How close is the value to what you predict?
23
Q
Sum of Cross Products
(Don’t need to know equation)
A
24
Q
Intercept of Regression
A
25
Q
Types of Correlation Coefficients
A
- Pearson’s
- Bi-serial R
- Point biserial R
- Tetrochoric R
- Phi
- Use depends on if data is continuous, dichotomous (artifical/true)
26
Q
Residuals
A
- Residuals: How wrong you are. Residual = 0 is perfect correlation
27
Q
Standard Error of Estimate
A
28
Q
Coefficient of Determination and Alienation
A
- Variability around a line
- What percentage of variation is being accounted for.
- If you didn’t have x and wanted to predict y, you should always predict the mean
- You can account for variance if you have x, which is r2
- Coefficient of alienation = to what extent are 2 variables not associated with each other (inverse of r2)
29
Q
Multivariate Models:
Linear Combination
A
- Examines more than one variable influencing a result
30
Q
Factor Analysis
(And steps)
A
- Know how to read a factor anlysis table and understand it
- Trying to find common factors among a large amount of variables or measurements
- Can make as many factors as there are items, but scree plot determines how many factors you should using point infleciton
- Can then get r2 for any of these factors to explain what they account for
- Find (orthogonal) lines through a cloud of data that explains the most variance
- Each line defined in a linear combination
- For each item see how highly it correlates with that linear combination
- Define the factors
31
Q
A
