20.10.2015 Views

A COMPENDIUM OF SCALES for use in the SCHOLARSHIP OF TEACHING AND LEARNING

compscalesstl

compscalesstl

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

eflect overall teacher behaviors or separated <strong>in</strong>to two subscales (or two factors) represent<strong>in</strong>g<br />

car<strong>in</strong>g/supportive teach<strong>in</strong>g behaviors and professional competency/communication skills.<br />

Researchers also <strong>use</strong> factor analysis <strong>for</strong> reduction purposes, <strong>in</strong>clud<strong>in</strong>g limit<strong>in</strong>g <strong>the</strong> number of<br />

items <strong>in</strong> an <strong>in</strong>strument as well as <strong>the</strong> number of factors <strong>in</strong>cluded <strong>in</strong> an <strong>in</strong>strument (Yong &<br />

Pearce, 2013). For example, ano<strong>the</strong>r scale discussed fur<strong>the</strong>r <strong>in</strong> Chapter 11 of this e-book is <strong>the</strong><br />

Satisfaction with Life Scale (SWLS) (Diener, Emmons, Larsen, & Griff<strong>in</strong>, 1985; see Layous,<br />

Nelson, & Legg, 2015). These authors reviewed <strong>the</strong> literature, found preexist<strong>in</strong>g scales of life<br />

satisfaction to be <strong>in</strong>adequate, and <strong>the</strong>re<strong>for</strong>e, began <strong>the</strong> process of develop<strong>in</strong>g a new<br />

<strong>in</strong>strument. In <strong>the</strong> first phase, <strong>the</strong>y developed 48 items related to satisfaction with one’s life,<br />

which was decreased to 5 items by us<strong>in</strong>g factor analysis. Procedures such as <strong>the</strong>se help<br />

researchers to f<strong>in</strong>d ways to more effectively measure a construct, while also reduc<strong>in</strong>g<br />

participant fatigue with answer<strong>in</strong>g numerous items and <strong>in</strong>creas<strong>in</strong>g <strong>the</strong> likelihood of replicat<strong>in</strong>g<br />

f<strong>in</strong>d<strong>in</strong>gs (Henson & Roberts, 2006).<br />

It is also important to note that statistical programs will only show where and how items load; it<br />

is up to <strong>the</strong> researcher to <strong>in</strong>terpret <strong>the</strong> f<strong>in</strong>d<strong>in</strong>gs us<strong>in</strong>g <strong>the</strong> literature and/or a <strong>the</strong>oretical<br />

framework (Henson & Roberts, 2006; Yong & Pearce, 2013). For example, delet<strong>in</strong>g items as<br />

suggested by SPSS does not always <strong>in</strong>crease reliability (Cho & Kim, 2015). Depend<strong>in</strong>g on <strong>the</strong><br />

f<strong>in</strong>d<strong>in</strong>gs of <strong>the</strong> factor analysis, researchers might go back to <strong>the</strong> literature and develop<br />

additional items to pilot or determ<strong>in</strong>e <strong>the</strong> <strong>in</strong>strument’s structure and assess its reliability and<br />

validity (Warner, 2013).<br />

Reliability<br />

Regardless of whe<strong>the</strong>r an <strong>in</strong>strument be<strong>in</strong>g <strong>use</strong>d <strong>for</strong> SoTL is preexist<strong>in</strong>g, an adapted measure, or<br />

newly developed, it is important to assess its reliability. Generally speak<strong>in</strong>g, reliability refers to<br />

how consistently an <strong>in</strong>strument measures a certa<strong>in</strong> construct or variable of <strong>in</strong>terest (John &<br />

Benet-Mart<strong>in</strong>ez, 2000). Scores on <strong>the</strong> <strong>in</strong>strument should not change unless <strong>the</strong>re is a change <strong>in</strong><br />

<strong>the</strong> variable. It is not possible to calculate reliability exactly, as <strong>the</strong>re is <strong>in</strong>evitable measurement<br />

error. There<strong>for</strong>e, researchers have to estimate reliability. In do<strong>in</strong>g so, all attempts to remove or<br />

control sources of error should be made so as to obta<strong>in</strong> a truer score of <strong>the</strong> latent variable<br />

(DeVellis, 2012).<br />

The question <strong>the</strong>n becomes, how do we know if an <strong>in</strong>strument is reliable? The methods <strong>for</strong><br />

assess<strong>in</strong>g reliability can vary depend<strong>in</strong>g on type of data and <strong>in</strong><strong>for</strong>mation sought (John & Benet<br />

Mart<strong>in</strong>ez, 2000); however, <strong>the</strong>y are all based on <strong>the</strong> idea that reliability is <strong>the</strong> proportion of<br />

variance of <strong>the</strong> observed score that can be attributed to <strong>the</strong> true score of <strong>the</strong> latent variable<br />

be<strong>in</strong>g exam<strong>in</strong>ed (DeVillis, 2012). Many researchers will report a s<strong>in</strong>gle coefficient alpha<br />

regard<strong>in</strong>g reliability (Hogan, Benjam<strong>in</strong>, & Brez<strong>in</strong>ski, 2000), but that is a limited picture of<br />

reliability and not always <strong>the</strong> best <strong>in</strong>dicator (Cho & Kim, 2015; Cronbach & Shavelson, 2004;<br />

John & Benet-Mart<strong>in</strong>ez, 2000). For example, Cort<strong>in</strong>a (1993) demonstrated how alpha can<br />

<strong>in</strong>crease with <strong>the</strong> number of items <strong>in</strong> an <strong>in</strong>strument despite low average item <strong>in</strong>tercorrelations.<br />

There<strong>for</strong>e, it is important to exam<strong>in</strong>e whe<strong>the</strong>r additional reports of item relationships and/or<br />

reliability are given (Schmitt, 1996). There is also somewhat of a challenge <strong>in</strong> determ<strong>in</strong><strong>in</strong>g<br />

35

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!