# Correlation Between Changes

### Richard Chambers

Measures of change were derived using linear regression (see Method), with the pre-score (T1) used to predict the post-score (T2). These analyses were used to derive a standardised residual score for each dependent variable which represented the change in that variable over the intervention period. These standardised scores were then analysed for correlation with change residuals from other variables in order to evaluate which changes covaried over the intervention period.

As predicted in Hypothesis 6, statistical analyses indicated a number of significant correlations between (1) group membership, (2) the degree of change between self report measures of affect, (3) the degree of change between self report measures of metacognitive processing, and (4) direct measures of executive function. For example, as predicted in Hypotheses 1, 2, and 4, (and consistent with the ANOVA analyses presented above) group membership correlated significantly with improved scores on the BDI (r = .45, p = .004), the negative affect dimension of the PANAS (r = -.32, p = .045), the MAAS (r = -.50, p = .001), the DSB (r = -.41, p = .008), and the neutral condition of the IST (r = .37, p = .02). Correlations between change on measures of affect, metacognitive processing, and executive function are presented in Table 4.

#### Table 4 Correlations Between Measures of Affect, Metacognitive Processing, and Executive Function

Measure 1 2 3 4 5 6 7
1.BDI -

2.BAI .44** -

3.PANAS + -.38* -.19 -

4.PANAS - .1 -.1 .01 -

5.RRS .28* .23 -.36* .04 -

6.MAAS -.49** -.28 .37* -.12 -.36* -
7.DSB -.18 -.24 .22 -.09 -.1 .31 -
8.IST .39** -.05 .04 .12 .08 -.15 -.005

Note: correlation of switch-cost residuals with other variables was calculated for N=39, since one case was removed during data screening as a multivariate outlier. * p < .05 ** p < .01.

Self-Report Measures of Affect. Decreased scores on the BDI correlated significantly with decreased scores on the BAI (r = .44, p = .002), and, negatively, with change on the positive affect dimension of the PANAS (r = -.38, p = .009), as would be expected, since measures of affect are generally highly correlated. Interestingly, though, the degree of change on the negative affect dimension of the PANAS did not significantly correlate with change on any other self-report affect measure.

Measures of Metacognitive Processing. As predicted in Hypothesis 3, reduced scores on the RRS correlated significantly with increased scores on the MAAS (r = -.36, p = .01).

Measures of Executive Function. Since it was hypothesised that the meditation course would enhance attention-switching capacity and result in reduced switch-trial RTs in the meditators as compared to the controls (Hypothesis 4), only RTs for the switch trials were included in these analyses.

There was an overall correlation between change in switch-condition RTs, in the Neutral and Affective conditions, of r = .49, p = .001.

Self-Report Measures of Affect and Self-Report Measures of Metacognitive Processing. As predicted in Hypothesis 3a, improved scores on the RRS correlated significantly with improved scores on the BDI (r = .28, p = .04) and the positive affect dimension of the PANAS (r = -.36, p = .03). Likewise, improved scores on the MAAS correlated highly significantly with improved scores on the BDI (r = -.49, p = .001).

Self-Report Measures of Affect and Measures of Executive Function. As predicted in Hypothesis 6, improved scores on the BDI correlated highly significantly with reduced RTs on the IST (r = .41, p = .009). Further analysis indicated that this was due to a highly significant correlation between improved BDI scores and reduced RTs in the Neutral condition (r = .43, p = .007). There was a trend toward a significant correlation between improved BDI scored and improved RTs in the Affective switch task (r = .28, p = .04).

Self-Report Measures of Metacognitive Processing and Executive Function. There were no significant correlations between improved scores on the RRS and MAAS and decreased residual scores on the IST.

As predicted in Hypothesis 6, improved scores on the DSB correlated highly significantly with improved scores on the MAAS (r = .314, p = .03).