Today

  • Course Overview
  • Statistical review (continued)
  • Brief history of testing
  • Norms and sampling

About the Course

  • E-411-PRMA Course Website
  • Cohen, R. J., Swerdlik, M. E., & Sturman, E. D. (2013). Psychological testing and assessment: An introduction to tests and measurement. New York, NY: McGraw-Hill
  • Website readings on item response theory and generalizability theory
  • Will be using R in class and on assignments
  • http://kennslubanki.hi.is/tolfraedi

First-Day Test

  • Thoughts?
  • What was I measuring?
  • Pros/cons of the tests?
  • Will return to this test throughout the course

Testing and measurement

  • What is measurement?
    • Assignment of numerical values based on a set of rules
  • What is a test?
    • Instrument used to measure

Statistical Review

  • Scales - Nominal, ordinal, interval, and ratio
  • An item measuring ...
    • Temperature (in Celsius)
    • Temperature (in Kelvin)
    • Speed
    • Gender
    • Happiness

Probability distribution, Skewness, kurtosis

  • What is a probability distribution
    • Assigns a probability, likeliness of occurence, of a score of all possible scores
    • May be parametric or non-parametric
    • Normal
  • What skew might you expect these outcomes to look like?
    • Reaction time in a psychological experiment
    • Number of children in a family
    • Scores on an easy test
    • Height in Reykjavík
  • Platykurtic, mesokurtic, and leptokurtic
  • Plot your data, rely less on statistics!

Homework Problems

  • GRE exam: Mean = 500, SD = 100, and 1000 participants
  • Jarl gets a 650 on the GRE, about how many students scored below him?
  • 330 students score below Þöll, what was her score?

Letting R do the work

        pnorm(650, mean = 500, sd = 100) * 1000
        # 933.1928
        qnorm(330/1000, mean = 500, sd = 150)
        # 456.0087

Percentiles

  • 330 students score below Þöll, what percentile is she in?
  • Percentiles represent a reasonably simple and intuitive way to classify scores relative to other test takers

Correlation

$$ r = \frac{\sum (X - \bar{X})(Y - \bar{Y})}{\sqrt{\sum(X - \bar{X})^2 \sum(Y - \bar{Y})^2}}$$

Calculating correlations

XY
56
30
10
Mean32

Calculating the correlation in R

        # Assign X the values of 5, 3, 1 and Y the values of 6, 0, and 0
        x <- c(5, 3, 1)
        y <- c(6, 0, 0)
        # Calculate the correlation
        cor(x , y)
        # 0.8660254

Simple linear regression

  • Pretend we are interested in predicting height given someones weight
  • We might consider a linear regression model
$$Height = \beta_0 + \beta_1*Weight$$
How would we know if this is appropriate?
Description

Model Summary

CoefficientEstimate
Intercept25.723456
Slope0.287249
$$\hat{Height} = 25.723456 + 0.287249*Weight$$
How does this relate to the correlation?

Slope and the correlation

  • There is a relationship between the estimated slope and the correlation between two variables in a SLR
$$ r = b\frac{SD_X}{SD_Y} $$
  • What is r if b is 0.287249, the SD of weight is 15.49869 and the SD of height is 4.472136?
  • Does that parallel what you thought based on the plot?
  • What would b become if r = 0.4?

History and timeline of test development

  • 2200 BCE, Chinese believed to use testing for determining who would get governmental jobs
  • Greek and Romans categorized individuals based on personality type ("blood" or "phlegm")
  • Francis Galton's classification based on "natural gift" (i.e. eugenics)
    • Contributed to development of questionnaries, rating scales, and self-report inventories
  • Wilhelm Wundt's laboratory and his focus on "standardization"
    • James Cattell's mental tests
    • Charles Spearman - reliability and factor analysis

Testing in the 20th Century

  • 1905, Binet and Simon publish a test measuring intelligence in mental retarded school children in Paris
  • 1939, Wechsler publishes a test to measure intelligence in adults (would become WAIS)
  • Group intelligence test administered by the US military during WWI and WWII
  • WWI personality tests used to screen recruits

Assumptions of Psychological Testing

  • Psychological traits and states exist
  • Psychological traits and states can be measured
  • Behavior on tests predicts non-test behavior
  • Measurement error is part of the process
  • Test can be fair
  • Test can benefit society

Norm-Referenced Testing

  • Individuals scores are relative only to some reference group
  • This group should represent the entire pool of test takers for the tested construct
  • Collectively, this group is known as a normative sample and data from them are the norms
  • Understanding the normative sample is very important

Sampling: PTSD in Iceland

Description

Sampling Techniques

  • Simple random sample
  • Stratified random sample
  • Cluster random sample
  • Purposive sample
  • Convenience sample

Different Norms

  • Developmental Norms
    • Age Norms
      • A 6 year old performs at the level of a 10 year old
      • This is on this material only though!
    • Grade Norms
      • School year typically 10 months in the US (and Iceland?)
      • A 4th grader is performing at the level of a 5th grader in third month
      • This is on this material only though!
  • National Norms, national representative
    • Anchor norms enable two tests to be compared
    • In USA, students could take SAT or ACT for admission to college

Fixed Reference and Criterion-Related

  • Fixed reference group scores are used as the basis for calculation of future administrations of the test
  • SAT does this through using anchor items and equating
  • Criterion-referenced, evaluate a score with reference to a set criteria or standard NOT other test takers

Question

  • How would you score the grades in a class room?
  • What do you think is the fairest way?

Next time

  • Please read the chapter on reliability (chapter 5)
  • Please watch the RStudio videos