Reliability Assessment with Skewed Data: To Transform or Not to Transform?

2.50
Hdl Handle:
http://hdl.handle.net/10755/150126
Type:
Presentation
Title:
Reliability Assessment with Skewed Data: To Transform or Not to Transform?
Abstract:
Reliability Assessment with Skewed Data: To Transform or Not to Transform?
Conference Sponsor:Sigma Theta Tau International
Conference Year:2003
Author:Norris, Anne E., RN, PhD, CS
P.I. Institution Name:Boston College
Title:Associate Professor
Co-Authors:Karen Aroian, RN, PhD, CS, FAAN
Objective: Although generally recommended, data transformation has hidden costs and sometimes, little benefit. This paper presents guidelines for evaluating degree of skew present, and case for and against transforming skewed data when assessing instrument reliability. Design: Descriptive, longitudinal Population & Setting: Immigrants (n=758) from former Soviet Union, living in Boston area and 50 years or older in age (mean age = 63 years; mean time in US = 4 years; 57% women). A small subsample (n=188) were interviewed twice to provide test-retest data. Concepts: Reliability, data transformation, skew. Methods: Participants were interviewed by Russian speaking interviewers who were themselves former immigrants from the Soviet Union. As part of this interview, participants completed a Russian language version of the SCL-90R developed through translation and back translation. Cronbach's alpha was calculated using original, square root transformed, and log transformed data. Test-retest was calculated using sums of original item responses for each time point, and square root and log transformations of these sums. Additionally, test-retest was calculated using sums of square root and log transformed items for each time point. Findings: 34% of items were moderately (skew > |1.25|), and 21% highly skewed (skew > |2.25|). However, neither Cronbach's alpha nor test-retest differed for original and transformed data, except when the sum of transformed items was used to calculate test-retest, for both total and small, randomly created subsamples. Sums of transformed items inconsistently resulted in slightly higher and lower values. Conclusion: Data transformation is not always needed or advisable when calculating Cronbach's alpha or test-retest reliability for an instrument with skewed item responses. Implications: Researchers should consider the statistical analysis that is planned on a case by case basis, as well as degree of skew present in data before transforming them. The best answer to "Should I transform these data?" is "It depends."
Repository Posting Date:
26-Oct-2011
Date of Publication:
17-Oct-2011
Sponsors:
Sigma Theta Tau International

Full metadata record

DC FieldValue Language
dc.typePresentationen_GB
dc.titleReliability Assessment with Skewed Data: To Transform or Not to Transform?en_GB
dc.identifier.urihttp://hdl.handle.net/10755/150126-
dc.description.abstract<table><tr><td colspan="2" class="item-title">Reliability Assessment with Skewed Data: To Transform or Not to Transform?</td></tr><tr class="item-sponsor"><td class="label">Conference Sponsor:</td><td class="value">Sigma Theta Tau International</td></tr><tr class="item-year"><td class="label">Conference Year:</td><td class="value">2003</td></tr><tr class="item-author"><td class="label">Author:</td><td class="value">Norris, Anne E., RN, PhD, CS</td></tr><tr class="item-institute"><td class="label">P.I. Institution Name:</td><td class="value">Boston College</td></tr><tr class="item-author-title"><td class="label">Title:</td><td class="value">Associate Professor</td></tr><tr class="item-email"><td class="label">Email:</td><td class="value">norrisa@bc.edu</td></tr><tr class="item-co-authors"><td class="label">Co-Authors:</td><td class="value">Karen Aroian, RN, PhD, CS, FAAN</td></tr><tr><td colspan="2" class="item-abstract">Objective: Although generally recommended, data transformation has hidden costs and sometimes, little benefit. This paper presents guidelines for evaluating degree of skew present, and case for and against transforming skewed data when assessing instrument reliability. Design: Descriptive, longitudinal Population &amp; Setting: Immigrants (n=758) from former Soviet Union, living in Boston area and 50 years or older in age (mean age = 63 years; mean time in US = 4 years; 57% women). A small subsample (n=188) were interviewed twice to provide test-retest data. Concepts: Reliability, data transformation, skew. Methods: Participants were interviewed by Russian speaking interviewers who were themselves former immigrants from the Soviet Union. As part of this interview, participants completed a Russian language version of the SCL-90R developed through translation and back translation. Cronbach's alpha was calculated using original, square root transformed, and log transformed data. Test-retest was calculated using sums of original item responses for each time point, and square root and log transformations of these sums. Additionally, test-retest was calculated using sums of square root and log transformed items for each time point. Findings: 34% of items were moderately (skew &gt; |1.25|), and 21% highly skewed (skew &gt; |2.25|). However, neither Cronbach's alpha nor test-retest differed for original and transformed data, except when the sum of transformed items was used to calculate test-retest, for both total and small, randomly created subsamples. Sums of transformed items inconsistently resulted in slightly higher and lower values. Conclusion: Data transformation is not always needed or advisable when calculating Cronbach's alpha or test-retest reliability for an instrument with skewed item responses. Implications: Researchers should consider the statistical analysis that is planned on a case by case basis, as well as degree of skew present in data before transforming them. The best answer to &quot;Should I transform these data?&quot; is &quot;It depends.&quot;</td></tr></table>en_GB
dc.date.available2011-10-26T10:17:05Z-
dc.date.issued2011-10-17en_GB
dc.date.accessioned2011-10-26T10:17:05Z-
dc.description.sponsorshipSigma Theta Tau Internationalen_GB
All Items in this repository are protected by copyright, with all rights reserved, unless otherwise indicated.