The psychometric evaluation of the PGS is presented in terms of its internal structure, variability, reliability (internal consistency), and construct validity (theoretical relationship to other measures).
Of the 317 participants who received the 15-item spirituality scale, complete data were available for 308 participants. The sample size recommended for exploratory factor analysis is a minimum of 5 (Hatcher, 1994
), or 10–15 subjects per item (Pett, Lackey, & Sullivan, 2003
). The sample size of 308 was well within these guidelines, at about 20 subjects per item.
Exploratory factor analysis of the PGS was conducted using SPSS 16.0.2. The appropriateness of factor analysis was examined first for the 15 items. The Kaiser–Meyer–Olkin measure was 0.89, supporting the adequacy of the sample for these items (Tabachnick & Fidell, 2007
). Bartlett’s test of sphericity, χ2
= 317) = 3,392, p
< .01, allowed the conclusion that all the items were not uncorrelated, and therefore, the correlation matrix might be factorable (Pett et al., 2003
). The first two Eigenvalues of the unrotated model were greater than 1 (7.35, 2.37), suggesting a potential two-factor solution according to the Kaiser criterion. In addition, examination of the elbow
of the scree plot indicated that the optimal number of factors was two. To avoid missing potentially important information in a third factor, both two- and three-factor models were selected, and results were compared. Each model was fit using principal axis factoring with oblique rotation. The pattern matrix for the two-factor model is presented in .
Pattern Matrix for the Two-Factor Model and Factor Properties (n = 317)
The two-factor model had high primary loadings for all 15 items, suggesting that 9 items measure one factor and 6 items measure the other (). The very low cross-loadings suggest the existence of two well-separated factors, and retention of all 15 items to measure these factors is supported. In the three-factor model, the 9 items loading heavily on the first factor are the same ones that loaded on the first factor in the twofactor model. Four of the remaining items loaded heavily on the second factor, and two loaded heavily on the third factor. The primary loadings in the three-factor model were high, but the cross-loadings were not as low as in the two-factor model. In addition, the third factor was supported primarily by only two items, and there was general agreement that a scale should have at least three items loading on any given factor (Tabachnick & Fidell, 2007
). Thus, the two-factor model was chosen. The two factors explained 59.4% of the total variance.
The first factor, Support From God, reflects the perspective of a direct connectedness to God. The emphasis of this subscale is on looking beyond self and less on the illness to the powerfulness of God. Thus, survivors have a personal relationship with God, are drawn to God through the illness, and take comfort in knowing that God is in control. The second factor, God’s Purpose for Me, reflects strategies used to cope with the earthly realities of illness. The emphasis of this scale is more on the illness and how God is working through the illness to build character within the self. In summary, in the first factor, attention is cast on God, and in the second factor, attention is cast on how the self is being reformed by God through the illness.
Subscales and Descriptive Statistics
The two subscale scores were computed by averaging responses on items with primary loadings on each factor. To allow computation of scores for participants with small amounts of missing data, subscale scores were computed if a participant was missing no more than one item on a factor, yielding 317 participants with a score for Support From God and 316 participants with a score for God’s Purpose for Me.
Descriptive statistics for the two PGS subscales, each with a possible range of 0 to 4, are shown in . The mean score for the factor Support From God was 3.58 (SD = 0.76), with a range of 0 to 4. The second factor, God’s Purpose for Me, had a mean score of 2.57 (SD = 1.19). The median scores for the two subscales were 4.00 and 2.73, respectively. The average scores for the factor Support From God were highly negatively skewed (−2.6) and had low variability; more than half of the participants scored at the upper limit of the subscale. Scores for the second factor, God’s Purpose for Me, had a slight negative skew of −.58, and the variability was higher. The correlation between the two factors was .48.
Item Analysis and Internal Consistency
Both PGS subscales had mean interitem correlations above .30. Within a subscale, no interitem correlation fell below .30. Most corrected item-total correlations fell within the recommended .30 to .70 range (Nunnally & Bernstein, 1994
). Cronbach’s α
coefficients were strong, .94 for the Support From God factor and .86 for the God’s Purpose for Me factor, both well above the .70 cutoff suggested by Nunnally and Bernstein (1994)
To evaluate test–retest reliability, the scale was administered to a convenient subsample of 19 participants 2–3 weeks later. Test–retest reliability obtained with Pearson’s correlations was .94 for the first factor, Support From God, and .88 for the second factor, God’s Purpose for Me. Even though the test–retest was based on only 19 participants, the reliabilities were quite high, supporting the temporal stability of the instrument.
Of the 317 Christian African American cancer survivors who completed the PGS, a subsample of 137 participants were administered the Religious Involvement Scale and 274 were administered the GDS-SF15. The average score on the GDS-SF15 was 3.3 (SD = 2.84), with 17% (n = 53) scoring above the cutoff, indicating possible depression. On the Religious Involvement Scale, the average score was 8.4 (SD = 3.75) on the organizational religiosity subscale, 15.8 (SD = 3.4) on the nonorganizational religiosity subscale, and 11.3 (SD = 1.19) on the subjective religiosity subscale. The construct validity of the PGS was calculated by computing correlation coefficients with the Religious Involvement Scale and the GDS-SF15 (). The demographic characteristics of these subsamples were very similar to those of the larger sample summarized in , so they are not reported separately.
Pearson Correlation Coefficients Relating the Factors of the Perspectives of Support From God Scale to Existing Scales
Validity evidence based on relations of the PGS to the three dimensions of religious involvement was moderate to strong, lending support for the first hypothesis. The subscale scores on Support From God and God’s Purpose for Me were correlated positively with all the Religious Involvement subscales (r = .34 to .62 for Support From God; r = .20 to .41 for God’s Purpose for Me). The PGS subscale that correlated most highly with religious involvement was the Support From God subscale, although the correlations with organizational religiosity were lower than for nonorganizational and subjective religiosity.
Support for the second hypothesis, that spirituality would correlate negatively to depression, was mixed. High scores on the PGS subscale Support From God was correlated negatively with depression (r = −.23, p < .01). However, the subscale scores on God’s Purpose for Me were not correlated significantly with depression.