Followers were asked to express their level of self-leadership behavior (n = 372). Data was gathered using the Self-Leadership Skills Inventory. The SLSI is a fairly new instrument developed by Furtner and Rauthmann (in prep.). The present study expands upon existing research on self-leadership by providing a surround validation of the SLSI. Prior to using the collected data for analysis, a confirmatory factor analysis is anticipated. Factor analysis of confirmatory nature should be pursued in order to develop and/or validate an instrument (Janssen & Laatz, 2013, p. 547). In the current work, items of the SLSI were clustered in nine blocks of each three statements, representing items belonging to different factors, in order to minimize the risk of bias due to response patterns. To test whether the data was suitable to proceed with factor analysis, two prior tests are carried out.

One test for sample data adequacy is the Kaiser-Meyer-Olkin (KMO) measure (Kaiser, 1970). The KMO provides the ratio of squared partial correlations and squared correlations between all variables in the model (Field, 2013, p. 684). Val?ues may vary between 0 and 1, where values close to 1 indicate appropriateness to carry out factor analysis. Values below 0.5 are deemed inacceptable (Janssen & Laatz, 2013, pp. 573-574). The second test applied is Bartlett's test of sphericity. It assesses whether correlation coefficients are significantly different from zero. The result of Bartlett’s test indicates whether the data is suitable for structure detection (Janssen & Laatz, 2013).

KMO and Bartlett’s test of sphericity were executed for 372 followers’ responses to the SLSI. Both indicators display satisfactory outcomes. KMO with a value of .87 is regarded as meritorious by Hutcheson and Sofroniou (1999). Bartlett’s test indicates significance of the correlation matrix being different to the identity matrix with x^{2} = 8048.18 at 351 degrees of freedom (p < .001). Due to the excellent adequacy measures of the data, a confirmatory factor analysis was pursued.

For the analysis of the SLSI, promax rotation with Kaiser-normalization and к = 4 was applied similar to earlier procedures by Furtner and Rauthmann (in prep.). Promax is an oblique rotation method, taking intercorrelation of factors into account (Janssen & Laatz, 2013, p. 568). Factor scores are reasonably high, reflecting values between 0.66 (item 7) and 0.97 (item 1). Furthermore, none of the items shows factor loadings equal or above 0.20 on other factors. The nine-factor structure proposed by Furtner & Rauthmann (in prep.) could thus be confirmed. The 27- item solution of the SLSI explains 83.59% of the variance of the measure. Outlined by Field (2013), the variance of the total-item solution should explain at least 50%.

Psychometric properties of the SLSI were computed using structural equation modeling. With a sample size of 372 followers indicating their self-leadership behavior, conditions for calculating a structural equation model are met (Weiber & Muhl- haus, 2014). The model was created using IBM SPSS AMOS 21 (Arbuckle, 2011). Various fit indices that are often used to explain model fit are determined. Calculating the relative chi-square (x^{2}/df) by Wheaton, Muthen, Alwin, and Summers (1977), the ratio should not exceed a value of 5.0. For root mean square error of approximation (RMSEA), indices of 0.01 show excellent, 0.05 good, and 0.08 mediocre model fit (MacCallum, Browne & Sugawara, 1996). Normed Fit Index (NFI), Goodness of Fit Index (GFI), Comparative Fit Index (CFI) should be greater than .90 (Byrne, 1994). For the Tucker-Lewis Index (TLI) results should exceed a value of .95 (Sharma, Mukherjee, Kumar & Dillon, 2005).

For followers’ self-assessment of self-leadership, model fit indices revealed acceptable estimates. The relative chi-square indicates good results with x^{2}/df = 3.00. Goodness of Fit Index (GFI = .85) and Normed Fit Index (NFI = .89) are slightly below the recommended amplitude of .90. Comparative Fit Index (.92), Tucker

Lewis Index (TLI = .91) as well as Root Mean Square Error of Approximation (RMSEA = .074) show satisfactory scores. These estimates declare the structural model to fit the data well.

Assessing highest-order factor structures, cognition-based strategies display the highest factor loading, (.96) compared to natural-reward strategies (.95) and social self-leadership strategies (.82). Single-item factor loadings of the SLSI range from .73 (item 7) to .95 (item 17). All first-order factors display stable factor loadings. Table 13 displays factor loadings of the SLSI calculated with IBM AMOS 21 (Ar- buckle, 2011).

Table 13. Factor Analysis of the SLSI with Promax Rotation

Item

Scale

I II

III IV V VI

VII VIII IX

20

PF

.90

23

PF

.93

26

PF

.84

10

GO

.88

13

GO

.85

16

GO

.80

2

SV

.90

5

SV

.94

8

SV

.85

1

SR

.78

4

SR

.83

7

SR

.73

12

PR

.79

15

PR

.90

18

PR

.81

3

SE

.83

6

SE

.91

9

SE

.91

11

IN

.83

14

IN

.87

17

IN

.95

19

SP

.84

22

SP

.87

25

SP

.86

21

SA

.85

24

SA

.83

27

SA

.88

Note.

n = 372, PF

= positive focus,

GO = group optimization, SV

= self-verbalization, SR = self-

reminding, PR = performance referencing, SE = success envision, In = intrinsification, SP = strategic planning, SA = self-analysis. Rotation method: promax (к = 4) with Kaiser-normalization.

High Cronbach alpha values (Table 14) and reasonably high factor loadings provide support for the development of the scale.

The global self-leadership mean score is reported at 3.33 (SD = 0.67). Higher-order factors of cognition-based strategies (M = 3.20, SD = 0.73), natural reward strategies (M = 3.19, SD = 0.81), and social self-leadership strategies (M = 3.82, SD = 0.73) result in moderate to high values for all self-leadership dimensions. Mean values of first-order factors range from 2.67 to 3.86. Out of all subfacets, selfverbalization (M = 2.67, SD = 1.16) revealed the lowest, whereas performance referencing (M = 3.86, SD = 0.86) displayed the highest mean value.

Cronbach’s alpha scores were computed for all factors contributing to the SLSI. The global scale projected a coefficient alpha of .94 which indicates excellent fit of internal consistency. Cognition-based strategies (a = .89), natural reward strategies (a = .90), and social self-leadership strategies (a = .88) further provide good results. The entire set of first-order factors produced good internal consistency ranging from a = .82 for self-reminding to a = .93 for self-verbalization.

Table 14. Scale Statistics for Self-Leadership

Scales

Descriptives

Scale statistics

n

M

SD

Skewness

Kurto-

sis

a

Mean inter-item correlation

Global self-leadership

372

3.33

0.67

-0.33

0.28

.94

.38

Cognition-based strategies

372

3.20

0.73

-0.11

-0.15

.89

.41

Self-analysis

372

3.28

0.92

-0.34

-0.27

.89

.73

Strategic planning

372

3.67

0.82

-0.72

0.70

.89

.73

Self-verbalization

372

2.67

1.16

0.19

-0.96

.93

.81

Self-reminding

372

3.18

1.00

-0.11

-0.75

.82

.61

Natural reward strategies

372

3.19

0.81

-0.21

-0.21

.90

.51

Positive focus

372

3.30

0.92

-0.37

-0.03

.92

.79

Intrinsification

372

3.15

0.94

-0.28

-0.41

.91

.77

Success envision

372

3.12

1.10

-0.23

-0.73

.92

.78

Social self-leadership strategies

372

3.82

0.73

-0.80

1.05

.88

.54

Group optimization

372

3.78

0.79

-0.69

0.59

.88

.71

Performance referencing

372

3.86

0.86

-0.84

0.96

.87

.64

Note. Standard error of skewness = .126. Standard error of kurtosis = .252.

Despite high reliability values, one point of criticism is that the SLSI may suffer from low content validity as mean inter-item correlations are higher than recommended. All subscales outline values between .61 and .81. Mean inter-item correlations for the higher-order factors of natural reward strategies (.51) and social selfleadership strategies (.54) are further slightly greater than the suggested value of .50. Medium interrelatedness may be discovered on the global scale (.38). Table 14 illustrates descriptive statistics, reliability, as well as item-interrelatedness.