Title | Testing PSY 434 |
---|---|
Author | Gianna Bozzi-Rich |
Course | Psychological Testing |
Institution | University of Rhode Island |
Pages | 7 |
File Size | 95.3 KB |
File Type | |
Total Downloads | 16 |
Total Views | 131 |
All lecture notes ...
Testing Psych 434 MWF 10-10:50 9/16 Monday FAPE, LRE & IEP Detailed testing to see what criteria children met whether they need special educational services- had to provided in LRE. IEP- Educational and other services like special psychological services If a state decided not to comply with PL 94-142 the govt couldn't do much about it but they could withhold federal money from school districts - They were not providing that much anyways (ie: New Mexico told the govt to keep their money) Law Section 504- Civil rights law for disabled people - Its a civil rights crime to discriminate by disability - Now states like NM couldn't discriminate bc it was made illegal - ADHD children weren't eligible for FAPE and IEp before passage of 504 - Now they are eligible for services- It would be illegal to withhold services from children with ADHD bc of 504 Early Intervention Programs (IEP)- The earlier you provide services, the less intense it has to be later on- it will help their conditions Hypothetical Constructs/ Latent (hidden) Traits: Intelligence- IQ tests which isn't able to be seen heard or touch. Instead we infer from behavior - Collect sample of behavior and what it represents for the latent trait Intra individual comparison- comparing within the individual Inter individual comparison- comparing between people True/False Test- Social Desirability Responding- People want to respond how they think they should/ what people want to hear How to reduce SDR- make test anonymous
9/23 Monday Standard Z scores for std deviation (normal) Percentiles: Used when tests are given to groups Associated with normal curve When closer to mean on normal curve, the percentiles are very close together and vice versa Ordinal scale for percentiles on the curve - Write out what central limit theorem is for extra credit Grade equivalence: - Interpolation - Extrapolation: extending line on graph beyond the point They are scaled in ordinal units You can’t subtract ordinal data- also can’t tell what std dev is so you can’t assume a child is behind a year If the std dev is one month then it is larger and not in the avg area of the curve False standard of comparison- not knowing std dev and trying to subtract ordinal scale score Age equiv: chronological age of person not grade level (12-3) 3rd month of being 12 years old and grade equiv: (3-1) 1st month of first grade ] 9/25 Wednesday Factor analysis: how do items intercorrelate with each other? Exploratory factor analysis: - Choose a method and create a matrix - Determine number of factors - Rotate factors - Associations between factors and terms - Associations between factors
Study for midterm: - History - Transforming scores - Ethics Eigenvalues- are related to proportions of variance - We develop a Scree Plot from eigenvalues - Scree plot: vertical axis is the eigenvalue / horizontal axis is the number of factors - Get data for plot from the table - Choose the number right before it levels off - Z scores have a mean of 0 and std dev of 1 T scores have a mean of 50 and std dev of 10 - Page 11-13 participant reactivity - Standardization - Reliability - Validity - Scores characteristics - Scales of measurement Orthangal rotation- factors are independent of each other which is different from oblique rotation where factors are correlated with each other Eigenvalue- how many factors are in the data set. Determine # of factors by plotting them against # of factors on scree plot Speeded tests are times and items are generally easy but too many to finish- measuring how many items correct in time allotted Present info in easily understood way directed to recipient - Individualize test according to person when going over results Multiple choice- to assess latent trait or hypothetical construct Percentiles- Variance = (std dev)^2
Friday 10/11 Regression line or line of prediction - If there is an x score and the relationship of x and y is represented in scatter, what Y score would be predicted? - Look at prediction line.. Not 100% accurate - If the correlation coefficient (+1) was perfect then every single point would fall on the line, meaning the prediction line would be 100% accurate - Linear relationship Pear shape- - Heteroscedasticity: y scores are distributed differently - The strength of the correlation coefficient is strong when close to the line - This varies in a pear shape Range- measure of variability Ie: variance, std deviation The higher the SD the larger the range, the more variability, the higher the correlation is going to be - Higher variability, higher correlation - Truncated range: restricted range - When we restrict the variance in one set of scores, we restrict or truncate the range - Truncated by preselection - WRAT test: wide range achievement test - Moderator variable: Monday 10/21 - When we have speeded tests (timed) you shouldn't use an internal reliability coefficient bc they're easy items but assessing how many can get correct Wednesday 11/6 A: Same trait, same method Monotrait, monomethod RELIABILITY Homotrait, homomethod
B: Same trait, different method Homotrait, heteromethod CONVERGENT VALIDITY Monotrait, heteromethod C: Different trait, same method Heterotrait, homotrait DISCRIMINANT/ DIVERGENT VALIDITY (Same method) Heterotrait, homotrait We DO NOT want high C count in multitrait multimethod matrix!!! D: Different trait, different method DISCRIMINANT/ DIVERGENT (Different method) Heterotait, heteromethod *chart on Sakai If we have true construct validity, then A will be the highest followed by B next C and then D…. A/B should be high and C/D should be low 11/13 Wednesday - Standard error of estimate= validity - SEM vs SE est - Rxy= + strong 11/22 Friday - High SES vs low ses - One regression line= no bias - Slopes is how steep the line is - Intercept is where the x and y connect - Under and over-predicting (line B) when using common regression line - Equal slope, but different intercepts= constant bias/ - Different slopes but same intercept= - Determining criterion related bias
Content validity 12/4 Wednesday - Item characteristic curve (ICC) - P level= difficulty in psychometrics or proportion of people who passed - What is the difficulty level of an item for a certain person - Graph of P values vs theta - Theta is a hypothetical construct (for example: how much a student learned in the course) - The slope tells us about the discriminability of an item - A steep slope means that the item is discriminating among the finer levels of theta - Tells us theta (trait), level of difficulty for level of trait (P), and steepness of the slope for discriminability ( how well an item differentiates among different levels) - Icc also tells us that - Correction for guessing- better to skip then try to guess because they subtract # wrong from # right (ie: GRE exam) - level of difficulty vs discriminability - Of everyone passes an item or everyone fails it, there is no discriminability - Best level of discrimination is P value thats in the middle of 0 to 1 = .5 (U shaped curve) 12/9/19 Monday...Review! Sensitivity: a/a+b Specificity : B/D+b ICC- parameters Horiz (x) : theta Vertical (y): difficulty level for corresponding level of theta Steepness of curve is discriminability Difficulty of the item.. The P of .5 level corresponds to what level of theta Where is the asymptote (.5 = true/ false item) - Can determine Guessing, discriminability, difficulty **No repeating of past questions but similar items from past exams Factor analysis- construct validity Divergent validity- low correlations between theoretically unrelated traits Degree of accuracy of a prediction(criterion predicted from the test score) - standard error of estimate How much error in test score- standard error of measurement
...