Advanced Educational Research: Review of Basics
Activity 3: Internal Consistency
The purpose of this exercise is to develop proficiency with assessing internal consistency of a set of items. Below is a link to responses from 19 students who completed the “Dissertation Process Survey.”
Note that there are 10 items on this questionnaire. The even numbered items (2, 4, 6, 8, and 10) were designed to measure anxiety toward the dissertation process. The odd numbered items were designed to measure self-efficacy toward the dissertation process.
What to do:
Using whichever statistical software is available:
(a) Calculate the internal consistency for anxiety items, and scale statistics (e.g., alpha if item removed, item-total correlations).
(b) Examine these items (carefully review wording of each item and examine item analysis statistics) and determine if any should be dropped. Explain why you dropped the item or items that you dropped.
(c) If some are dropped, recalculate internal consistency and re-examine the remaining items.
(d) Repeat these steps as necessary until the final sub-set of the items is obtained. These should be the best performing items (those that maximize alpha, or at least provide reasonable level for alpha). Once the final set of items is obtained, explain any items were removed, and interpret the alpha that was finally obtained.
(e) Repeat the process described above for the self-efficacy items.
What to submit for this exercise:
Nothing to submit; answers are provided. Compare your answers with those provided.
Let me know if you have questions.
Do not copy and paste tables from the statistical software you used (i.e., SPSS, Minitab, etc.).
Do not include any attachments.
1. Cronbach's alpha results
Simply summarize your findings directly and post findings on the Discussion Board in GeorgiaView, like this:
Five anxiety items (items 2, 4, 6, 8, and 10) produced alpha = .33. Removing item 2 would increase alpha to .53. Item 2 had a corrected-item total correlation of only .19, which is very low. Item 2 will be removed and the others will be retained since item 2 appears to be poorly matched with the others.
Item 2 removed, so alpha is .53 for items 4, 6, 8, and 10. Item 10 demonstrates low correct-item correlation of only .35 and alpha, if item 10 removed, would increase to .83. Etc.
Post results similar to format listed above.
2. Correlation results
Given the items selected from step 1 above, form two composite scores, one for anxiety and the other for self-efficacy, for each respondent. Using these composite scores, calculate Pearson's correlation and report results (using alpha = .05 significance level). About two sentences will be needed, for example:
There is a statistically significant, positive correlation between anxiety and self-efficacy toward the dissertation process (r = .32, p<.05). Results show that those who tend to have more self-efficacy -- confidence in their ability to handle the dissertation process -- tend also to have more anxiety toward the dissertation process.
3. t-test results
Using the composite scores created above, perform t-tests comparing males to females on both self-efficacy and anxiety and report results (using alpha = .05 significance level). For example:
There is no statistical difference in self-efficacy between males and females (t = -0.32, p>.05), but there is a statistically significant mean difference in anxiety scores (t = -3.34, p<.05) with females demonstrating greater anxiety.
Let me know if you have any questions.
Copyright © 2005, Bryan W. Griffin
Last revised on 18 February, 2015 01:21 PM