SDLRS: 요인분석 (Med Educ, 2005)

SDLRS: 요인분석 (Med Educ, 2005)

The Self-Directed Learning Readiness Scale: a factor analysis study

J Dennis Hoban, Sonya R Lawson, Paul E Mazmanian, Al M Best & Hugo R Seibel

의학에서 SDL은 중요하다. ABMS, RCPSC, WFME 등은 평생학습을 수련기간동안 반드시 평가해야 하는 전문성 중 하나로 보았다.

Self-directed learning is important in medicine where knowledge is continuously changing and a wide- range of patient problems is a constant.1 The American Board of Medical Specialties, The Royal College of Physicians and Surgeons of Canada, and the World Federation for Medical Education describe life-long and self-directed learning as professional characteristics that should be evaluated in the train- ing of physicians.2–4

1970년대 후반, Guglielmino가 SDL에 대한 준비도를 측정하는 도구를 개발하였다. Delphi 연구를 통해서 타당도를 마련했으며, 이러한 귀납적 접근법을 통해서 SDL에 대한 준비도를 개념화하고 다음과 같은 정의를 내렸다.

In the late 1970s, Guglielmino developed an instru- ment to measure self-directed learning readiness.5 A Delphi study was conducted to gain expert consensus on the characteristics of the self-directed learner. In the case of the SDLRS a foundational effort to ensure content validity involved basing the items on a Delphi panel s perception of the charac- teristics of an individual with a high level of readiness for self direction in learning’6 (p. 213). Using this inductive approach to conceptualise self-directed learning readiness, she offered a definition of the self-directed learner: 

• one who exhibits initiative, independence, and persistence in learning; 
• one who accepts responsi- bility for his or her own learning and views problems as challenges, obstacles; 
• one who is capable of self-discipline and has a high degree of curiosity; 
• one who has a strong desire to learn or change and is self-confident; 
• one who is able to use basic study skills, organize his or her time and set an appropriate pace for learning, and to develop a plan for completing work; 
• one who enjoys learning and has a tendency to be goal-oriented.5 (p. 73). 

Guglielmino는 이러한 특질과 관련된 문항을 개발해서 41문항으로 된 원본을 만들었고 PCA w/ varimax rotation으로 8개 요인 구조를 도출했다.

Guglielmino developed items that related to these qualities and produced her original version of the SDLRS consisting of 41 items. She tested it on a group of 307 US high school juniors and seniors, college undergraduates, and adults in continuing education courses. She reported that principal com- ponent analysis (PCA) with varimax rotation yielded an 8-factor structure. She labeled these factors: 

(1) openness to learning opportunities; 

(2) self- concept as an effective learner; 

(3) initiative and independence in learning; 

(4) informed acceptance or responsibility for one’s own learning; 

(5) love of learning; 

(6) creativity; 

(7) future orientation; and 

(8) ability to use basic study skills and problem solving skills.5 Later she modified the SDLRS to include 58 items.6 

SDLRS는 널리 사용되긴 했지만 비판도 많았다. 우선 타당도에 대한 Field의 지적이 있었다. Bonham도 구인타당도에 대한 의문을 삼았다. SDLRS점수가 낮다는 것은 두 가지를 의미하는데, (1) 공부를 싫어하거나 (2) 다른 사람에 의해 지도받는 것을 좋아하거나. 

The SDLRS has been widely used but also criticised. Field questioned the validity of the scale as a measure of readiness for self-directed learning.7 Responding to Field’s criticism, Guglielmino wrote that readiness in the scale title is a measure of an individual s current level of readiness to engage in self-directed learning ‘‘with the implication that this level may change’’’8 (p. 236). Bonham also questioned the construct validity of the SDLRS, suggesting that low scores on the SDLRS could mean two things: a dislike for learning or one’s preference for his or her learning to be directed by another.9 Brockett and Hiemstra stated, …the evidence is rather convincing that early concerns raised about certain items of the scale are warranted 10 (p. 73). 

이후 연구에서 SDLRS가 실제로 측정하는 것에 무엇인가에 대한 결과가 엇갈렸다. 

Subsequent empirical work yielded mixed results regarding what the SDLRS actually measured. 

• Mourad and Torrance administered the 58 itemSDLRS to a random sample of 684 K-12 students enrolled in a programme for gifted children at the University of Georgia.11 Their PCA suggested an 8-factor model; they concluded, more studies are needed to validate the scale using different samples (p. 102). 
• Field’s SDLRS study involved 244 adult students in Sydney, Australia.7 Using common factor analysis he identified 4 factors but then concluded that the scale measures a construct that is homogeneous 7 (p. 138). The single construct was love and enthusiasm for learning. 
• Bligh12 conducted an SDLRS study with medical trainees (n ¼ 216) during their medical education preparation in the UK. His PCA yielded 3 major factors: enthusiasm for learning, positive self- concept as a learner, and orientation to learning. 

Field의 연구와 다르게 여기서 언급한 모든 연구는 PCA를 사용했다. PCA와 Factor analysis는 공통점이 있지만, 다른 결과를 내놓기도 한다. 따라서 Field의 분석을 다른 연구자들의 분석과 비교하기는 어렵다. 실제로 McCune은 Field의 연구에 대해서 'PCA와 CFA의 결과 차이를 알아야 한다'라고 지적했다. 연구자들은 EFA를 위해서 종종 PCA를 활용한다. 그러나 우리는 Field의 연구가 Exploratory 하다고 생각한다. 여기서 제시한 연구들은 SDLRS에 깔린 구조를 알아내고자 하는 것이지만 PCA는 그러한 목적의 분석방법이 아니다.

With the exception of Field’s study, all the investiga-tions reported herein employed PCA. While PCA andfactor analysis methods may bear some similarities, they tend to produce different results. Thus it is challenging to compare Field’s factor structure with the others’ component structures. Indeed McCune criticised Field along these lines when she wrote, 'Also, Field should realise that if Guglielmino used principal component analysis while he used commonfactor analysis, their results should differ 13' (p. 245).Researchers often use PCA when conducting explor-atory factor analysis (EFA). We believe even Field’s study was exploratory in nature though McCune offered a competing view when she wrote, He statesthat ‘‘eight factors are sought’’ which I assume is his attempt to portray his analysis as a confirmatory factor analysis 13 (p. 245). PCA reduces a large set of items into smaller components and accounts for all ofthe variance among the items, but the studies we reported using PCA were trying to identify underlying structures of the SDLRS. PCA is not designed for that purpose. Preacher and MacCallum explain the distinction between PCA and exploratory factor analysis.14 

PCA yields observable composite variables (com- ponents), which account for a mixture of common and unique variance (including random error). The distinction between common and unique sources of variance is not recognised in PCA, and no attempt is made to separate unique variance from the factors being extracted. Thus, components in PCA are conceptually and mathematically quite different from factors in EFA.14 (p. 20). 

West와 Bently는 CFA를 이용해서 SDLRS척도를 분석하였다. orthogonal solution은 부적절하다는 결론을 내렸다. 6개 요인을 밝혔다. 이 중 3번 요인이 reverse scored item으로만 거의 이뤄져있었다. 또한 1st-order factor가 포함될 가능성을 추측해냈다.

West and Bentley examined the underlying SDLRS measurement model using confirmatory factor ana- lysis (CFA).15 Their study was conducted with 439 K-12 Tennessee teachers and administrators. The analysis concluded that an orthogonal solution to the SDLRS measurement model was not adequate. More importantly, they reported that a highly correlated 6-factor model best described the underlying theory. The 6 factors were: (1) love of learning; (2) self- confidence as a learner; (3) openness to challenge; (4) inquisitive nature; (5) self-understanding; and (6) acceptance of responsibility for learning. Inter- estingly, factor 3 contained mostly reverse scored items; yet, the possibility of reverse scored methods variance was not reported. Finally, they conjectured that the first order factors could be subsumed under a single factor characterising a higher order structure.

이들 연구에서 무엇을 배울 수 있는가? 첫째, 거의 모든 탐색연구가 reverse score item으로만 이루어진 요인이 있을 가능성을 제시한다. 그러나 이 것에 대해서 충분히 설명한 연구는 없다. 둘째, 거의 모든 연구자들은 EFA를 활용했다. CFA는 1st order factor들 간의 관계를 설명하는 2nd order factor를 찾게 해준다. 

What did we learn from the literature? First, nearly every exploratory study concluded that there was a component ⁄ factor that consisted of reverse scored items including Guglielmino’s5 original analysis. These items included negative statements about a high self-directed learner or a positive statement about a low self-directed learner. Not 1 author fully explained this phenomenon. Second, all researchers used EFA techniques except West and Bentley who used CFA methods. CFA allows the investigator to specify a second order factor to account for relationships among first order factors. In addition, it provides the researcher with information for disen- tangling random and systematic (method) error variance.16 For example, reserves scored items may be investigated as a source of method variance.

SDLRS의 psychometric nature를 연구하기 위해서 다음과 같이 했다.

To study the psychometric nature of the SDLRS for entering medical students, we collected SDLRS data from 972 students and conducted an EFA study with half the data and confirmed the model it produced with a CFA study using the other half of the data. Preacher and MacCallum emphasised the need for making good decisions in the process of conducting exploratory factor analysis.14 They were particularly concerned about using PCA, retaining components with eigenvalues greater than one, and using varimax rotation; a bundle of procedures affectionately termed Little Jiffy .14 This potentially limited approach was used in most of the previous SDLRS factor analysis studies we reviewed. We followed Preacher and MacCallum’s guidelines14 in our factor analysis.

17개 문항이 reverse scored되었다. 

Seventeen of the items are reverse scored. According to Guglielmino, wordingwas reversed in some of the items to prevent the response set of acquiescence (agreeing with all theitems).5

EFA 방법

Exploratory factor analysis – method 

'만약 요인들의 관계를 모른다면, 서로 완전히 독립이라고 가정할 이유가 없다. 따라서 oblique rotation을 하는 것이 안전하다.

We used Principal Axis Factor Analysis SPSS 11.0 for Windows to extract factors. To decide what factors to retain we used the scree plot, results from previous studies, and our comfort with the extracted factors. We decided to use an oblique rotation (Promax) based upon the West and Bentley study and Preacher and MacCallum’s argument, …if the researcher does not know how the factors are related to each other, there is no reason to assume that they are completely independent. It is almost always safer to assume that there is not perfect independence, and to use oblique rotation instead of orthogonal rotation 14 (p. 26). Individual loadings of 0.30 or greater were used in the factor designation. Extracted factors were examined and named based on an analysis of the items loading on each factor. Cronbach a was used to estimate the internal consistency of the items consti- tuting a factor.

EFA 결과

Exploratory factor analysis – results

5번째 요인이 모두 reverse scored item이었다.

Interestingly, the fifth factor measured all reverse scored items. This phenomenon was also reported by Guglielmino5, Field7, and Mourad and Torrance.11 Guglielmino noted her PCA showed that all items in factor 1 of her original 41 item version were negative statements.5 She speculated that it is possible that the factor also includes an avoidance of agreement with negative statements 5 (p. 61). Field7 and Mourad and Torrance11 also reported a factor in which items were phrased so that they had to be reverse scored. Mourad and Torrance offered 2 explanations for this factor (1) an attitude toward negative state- ments and (2) preference for complex and ambi- tious situations 11 (p. 99). Field labelled the factor containing all negatively worded items as facility with negatively phrased items 7 and argued it was not related to readiness for self-directed learning.

반복적으로 나타난 reverse scored phenomenon

Because the reverse scored phenomenon has been repeatedly reported in the literature we decided it was worth analysing in the confirmatory phase of our study.

CFA 방법

Confirmatory factory analysis – method

Using LISREL 8.5417 a series of confirmatory factor analyses was performed to further examine the measurement model underlying the SDLRS. First, the 58 items were trimmed to 41 in order to obtain items that loaded on only one factor in the specified model. Items were retained if: (a) they had factor loadings ¼ 0.30; and (b) their secondary loadings were < 0.30. This resulted in the deletion of 17 items.

We hypothesised a series of three models (Models ¼ A, B, C) based on the logic provided by Anderson and Gerbing.21

Model A is a 4-factor confirmatory model that represents the substantive factors derived from the previous EFA.

Model B is a 5-factor confirmatory model that represents 4 correlated substantive factors and an orthogonal reverse coding method factor that loads on reverse scored items from all substantive factors.

Next, Model C, a confirmatory higher order model was developed and evaluated. This model includes a higher order factor that is presumed to account for the 4 first order factors examined in Model A, and like Model B, includes a reverse scoring factor.

CFA 결과

Confirmatory factor analysis – results

Guglielmino는 귀납적 접근법으로 SDLRS를 만들었다. 그녀는 Delphi를 활용했으며, SDL에 바람직하거나, 필요하거나, 필수적인 것을 고려해서 만들었다고 확실히 언급했다. 또한 상황적, 태도적 문항이 필요해서 Delphi와 문항의 1:1 대응은 어려웠다고 했다. 본 연구에서 SDLRS는 Guglielmino가 말한 상황적, 태도적 특징을 충분히 측정하지 못하고 있는 것으로 보인다.

Guglielmino used an inductive approach to develop the SDLRS. She made clear, Results of the Delphi survey were used as a guideline in the construction of items for the scale. Characteristics which emerged from the survey with a rating of desirable, necessary,or essential were considered for inclusion. 5 (p. 37). She explained A one-to-one correspon- dence between SDLRS items and characteristics selected by the Delphi survey was not possible, since situational and attitudinal items were desired. 5 (p. 38). In our study, the SDLRS instru- ment did not fully measure these characteristics orsituational and attitudinal constructs that Gugliel- mino specified. 

본 연구의 EFA에서 4개의 요인을 밝혔다. McCune은 요인분석이 샘플에 따라서 다른 결과를 보일 수 있다고 했다. 따라서 (특히 다른 연구에서 PCA를 활용했다는 점에서) 다른 집단이나 다른 의대생의 결과와 다른 것에 놀라지는 않았다.

Our EFA study produced 4 acceptable factors. SinceMcCune cautioned that factor analysis studies of thesame instrument could yield different results depending on the samples used,13 we were not surprised that our EFA factors varied somewhat from the structures reported for other medical students12 and other populations,5,7,11,15 especially since many of our reviewed studies used PCA.

여러 연구들이 reverse scored item으로만 구성된 요인이 있음을 밝힌 바 있는데, 그 이상에 대한 연구는 없었다. SDLRS는 17개의 그러한 문항이 있다. 점점 더 많은 연구들이 reverse coded item이 내적일관성신뢰도를 떨어뜨리며 factor structure의 해석을 어렵게 함을 보여준다.

While several studies5,7,11 noted the presence of a factor comprising all or mostly reverse scored items, we learned that the reverse coding method variance had not been considered or further explored. The SDLRS contains 17 items that are reverse scored. A growing body of evidence concludes that reverse- coded items may weaken the internal consistency reliability of test score22–24 and impair interpretation of the factor structure.25–27

Marsh는 부정형으로 서술된 문항의 효과를 없애는 가장 쉬운 방법으로 긍정형 문장만 사용하는 것을 제시했다. 이것이 측정 전문가의 권고사항에 반하는 것이긴 하나, reverse coded item과 관련된 문제에 따른 제약이 있는 것은 분명하다.

Marsh suggests that the easiest way to eliminate the effects of negatively worded items is to use only positively worded items… 29 (p. 817). Although this approach runs counter to the recommendations of measurement experts, the problem of method vari- ance associated with reverse coded items suggests a limitation of including reverse scored items in the SDLRS.

여기서 얻을 수 있는 결론은 Guglielmino가 척도를 개발하기 위해 노력한 것은 맞지만 그 SDL척도는 부족해보인다. 우리 연구에 따르면 SDLRS는 다음의 네 가지를 측정한다. 

What can we conclude from our study of the SDLRS? We acknowledge Guglielmino’s efforts to develop a practical instrument for measuring self-directed learning readiness. Her carefully constructed approach to generating the instrument appears appropriate; yet, the SDLRS apparently falls short of measuring characteristics that Guglielmino deter- mined were associated with self-directed learning. With our two samples, the SDLRS measured the respondents’ perceptions of how often they felt positively about: 

• (1) learning being a tool for life; 
• (2) their self-confidence in their abilities and skills for learning; 
• (3) taking responsibility for their own learning; and 
• (4) their curiosity. 

SDL을 잘 하는 학생이 이러한 특징을 많이 가진 것은 직관적으로 말이 되는 것으로 보이지만, 이러한 인식이 SDL행동을 유도한다고 믿을 근거 또한 부족하다. Guglielmino는 원래 Delphi를 통해서 SDL학습자의 특징을 찾아냈을 뿐이고, 이에 대한 어떤 이론적 기반도 제시하진 않았다.

While it makes intuitive sense that self-directed learners would perceive themselves as holding these characteristics in abundance, there is no reason to believe that these perceptions would predict self-directed learning behaviour. Guglielmino identified only characteris- tics of self-directed learners from her original Delphistudy; she offered no theory of self-directed learningor of readiness.

지난 25년간 SDL에 대한 연구와 컨퍼런스가 많았지만 SDLRS가 바뀌지는 않았다. 의학에서 SDL에 대해서 SDL에 영향을 주는 조건에 대한 것도 설명해야 한다. 신경생물학의 최근 연구를 보면 다음의 것들을 중요시하고 있다. 학생과 의사들은 단순히 학습에 대해서 새로운 기술을 습득하는 것으로 충분하지 않다. 많은 경우 자신과 자신의 일과 지속적 전문성 개발에 대한 생각을 다시 할 것을 요구받는다. 우리는 Baveye가 말한 것처럼 SDL를 단순히 SDL에 대한 인식을 측정하는 것에서 관찰가능한 SDL노력이 어느정도인가를 측정하는 것으로 바뀌어야 한다고 생각한다.

Even as conferences and books exploring self-directed learning proliferated during the past 25 years, there appears to be no change in the SDLRS. In considering self-directed learning in medicine, one must account for the current research as well as conditions that influence the performance of self-directed learning. Newer studies extend fromthe neurobiology of aging and the role of cognition in making practice chan- ges30 to the social psychology of the physician’s changing work environment, the importance of phy- sicians’ peers, and the accountability schemes and financial incentives built into medical practice.31 Students and practitioners are not being asked merely to take on new skills or to adjust their attitudes toward learning. Many are being asked to rethink the way they see themselves, their work, and their ongoing professional development. We agree with Baveye32,33 who suggests that the study of self-directed learning should be reoriented to an entirely new direction, away from simple measures of perceptions of self- directed learning to observed self-directed learning endeavours, apropos of the 21st century.

---

Med Educ. 2005 Apr;39(4):370-9.

The Self-Directed Learning Readiness Scale: a factor analysis study.

Hoban JD1, Lawson SR, Mazmanian PE, Best AM, Seibel HR.

Author information

• 1Virginia Commonwealth University School of Medicine, PO Box 980565, Richmond, VA 232-0565, USA. jdhoban@vcu.edu

Abstract

BACKGROUND:

The practice of medicine demands that its physician practitioners are self-directed, life-long learners. The Self-Directed LearningReadiness Scale (SDLRS) intends to measure adults' readiness to engage in self-directed learning.

PURPOSE:

The present study assesses the underlying factor structure of the SDLRS for a sample of entering medical students.

METHODS:

Over a period of 6 years, 972 first year medical students at the Virginia Commonwealth University School of Medicine completed the SDLRS. To summarise the inter-relationships among variables, a principal axis factor analysis with oblique rotation was used on the 58 SDLRS items. A series of confirmatory factor analyses using LISREL 8.54 was performed to further examine the measurement model underlying the SDLRS.

RESULTS:

A 4-factor confirmatory model representing 4 correlated substantive factors and a reverse coding method factor fits these data well.

CONCLUSIONS:

Medical educators should hold limited expectations of the SDLRS to measure medical students' readiness to engage in self-directed learning. The definitions and theoretical assumptions that inform readiness for self-directed learning should be reconsidered. Alternative approaches to studying self-directed learning should be explored.

Comment in

• Self-directed learning--the importance of concepts and contexts. [Med Educ. 2005]

PMID:

 

15813759

 

[PubMed - indexed for MEDLINE]