Karteikarten
Measurement Unit 7
Karten in diesem Stapel (28)
- 13. Imagine you do a study investigating the correlation between attachment to pets and fear of worms and you get the following results. Cronbach’s alpha was .75 for the attachment measure and .82 for the fear measure, and the correlation between these two variables was .43. What is your estimate for the true correlation if you had measures with perfect reliability?• True= observed/ sqrt(alpha a)(alpha)
- 24. In class, we discussed how reliability and validity can be influenced by having excessively homogenous items. The following three questions pertain to this issue. Explain what is meant by “homogeneous items” and give an example.-make all of your items the same
-if you make a questionnaire and ask the same questionnaire over and over the answers are likely to be the same
-dimensions of temperament scale
-parents rate temperament of child one scale is rythymnicity
-in general my child falls asleep at the same time
-when my child goes to sleep its usually about the same time
State how excessively homogenous items will influence Cronbach’s alpha and explain why.
-if you have scale with items that are correlated with each other the higher the coronachs alpha will be, which will be an over inflated estimate of reliability
State how excessively homogenous items will influence validity and explain why.
-all of the questions are the same so the scale would have high reliability but you are not capturing the full domain of the concept (so poor content validity) sacrificing validity - 35. List three situations where it would be inappropriate or impossible to calculate Cronbach’s alpha.• Speeded or timed tests
• Impossible to calculate alpha for a single item measure
• Casual versus effect indicators - 4Define casual indicators and define effect indicators.Casual: indicator itself is directly causing construct
•Measure of life stresses, how many stressful life events have you encountered
• Would not expect them to be correlated
Effect indicators: indicators are the effect of a construct
• I’m feeling worried and nervous, someone anxiety caused them to feel worried and nervous
• Construct is causing the response - 58. What is the formula for calculating xest?• xbar_o+(alpha)(x_o-xbar_o)
- 69. In class, we discussed a formula for calculating xest. What is the purpose of this formula?• Correcting scores for expected regression to mean
• Because you know instrument isn’t perfect the true score may actually be closer to the mean - 710. What is the formula you can use for estimating sdmeas (note: this question is asking for the formula that could be calculated in a real-life situation and not a formula that requires knowing a person’s true score).• (sd)sqrt(1-alpha)
- 811. Fill in the three missing values in the table below. Approximate 95% confidence intervals Reliability Interval will be plus or minus standard deviations +/- z-score
- 9At the beginning of treatment, Cookie Monster had a score of 25 on the Revised Oral Obsessions Scale (ROOS). At the end of treatment, his score dropped to 19. The questions below pertain to this scenario. What additional information would you need to determine if this represents a significant change based on the reliable change index?• You need to know the standard error of difference
- 10Explain the steps you would need to take if you were using the reliable change index to determine if this change was significant.• Take the difference score and divide it by the standard error measure of difference, then if your value is greater than 1.96, that indicated reliable difference and it is significant
- 1120. In class we discussed several different examples of types of assessment situations that involve the use of raters. List three of these examples.• Researcher observes participants completing a lie or recorded structured task, interaction or interview
• Clinician makes ratings after a structured clinical interview
• Ratings of home environments
• Ratings of written samples
• Ratings of animal behavior - 1221. What does it mean to say that items are treated as an interchangeable class of items?• Items could be raters, in which each person may be given a different set of raters (items)
• Any item could be given to any person and any person could have a different set of items
• doesn’t matter which pool of items a person gets - 1322. Imagine a researcher calculates a statistic to measure agreement between raters. How is this similar to reliability and how is this different from reliability?• It is not reliability, the more raters agree the higher the reliability, however agreement is not reliability, bc is reliability is the % of variance that is true. Agreement does not equal the % of variance that is true. Can be useful, and is related, but is not reliability.
- 1423. In what type of situation would it be most appropriate to use an intraclass correlation as an estimate of reliability?• Ratings from coders (raters or judges)
- 15Explained variance involves differences _______between people
- 16Error variance involves differences ____________within people
- 17Total variance is equal to ________ plus _________sum of squares explained
sum of squares error - 18The intraclass correlation is equal to ____ divided by _______ .sum of squares explained
sum of squares total - 19What statistic indicates the extent of agreement between raters after controlling for the expected rate of agreement due to chance?Kappa
- 20Kappa is a statistic that indicatesthe extent of agreement between raters after controlling for the expected rate of agreement due to chance.
- 21Imagine you are doing an item analysis for a set of questionnaire items that use Likert rating scales. List some basic descriptive statistics that would NOT be influenced by internal consistency and that would be important for your item analysis. Also, state what types of issues or problems would you be looking for when calculating these statistics.• Mean, SD, frequency
• Not influenced by internal consistency - 22Imagine you are doing an item analysis for a set of questionnaire items that use Likert rating scales. List some statistics that would be influenced by internal consistency and that could be useful in your item analysis.• Alpha
• Item-total correlations - 23What is the definition of “difficulty” for a dichotomous item?Difficulty level (often called “p”) is the percent passing the item
- 24Based on classical test theory, which difficulty level will produce the highest reliability?.50 produces highest reliability
- 2532. How would you calculate a discrimination index value for a dichotomous item?• Using total test scores, form upper & lower groups (e.g., top 25% and bottom 25%)
• (discrimination = (p for upper group) – (p for lower group) - 2633. Imagine you calculate an average inter-item correlation. How is this value similar to Cronbach’s alpha? How is it different from Cronbach’s alpha?• Extent to which there is correlation between your scales, high internal consistency = high alpha,
• Inter-item correlation is the average of the correlations b/w items
• Alpha as you have more items your alpha goes up
• Average inter item does not go up as items increase
• It is similar because it is related to internal consistency - 2734. What is a standardized alpha?• The alpha that would result if all items were turned to z-scores before summing.
- 2835. What would be the purpose of standardizing items one a composite scale before they are summed or averaged?• Standardizing items would make all items carry the same weight but is rarely worth the effort.
Informationen
- Beschreibung
- graduate measurement Karteikarten am Measurement Unit 7 , erstellt von abbi gutierrez am 14/03/2021.
- Erstellt von
- abbi gutierrez
- Aufrufe
- 6
- Bewertung