# Hierarchical Regression Analysis With Interaction Effects ## 4.3. Hierarchical Regression Analysis With Interaction Effects Initially, two three-step hierarchical regression analyses were conducted with the target variables overall empathy (QCAE) and overall mentalisation (MentS). Afterwards, QCAE’s Cognitive Empathy and Affective Empathy subscales, and MentS’ Self-Related and Other-Related subscales were selected as target variables as well. In the first step, the demographic variables age, sex, and monthly spending were modelled as control variables. In the second step, the main predictors of book count, overall existential concerns score, state anxiety, and trait anxiety scores were added. In the last step, the interaction between book count and the other main predictors were added. Regarding the control variables, monthly spending and education level were recoded into broader categories to address low cell sizes. Collapsed groupings and their respective sample sizes are shown in Table 7 below. | Collapsed Group | Original Group(s) | _n_ | % | | -------------------- | :-----------------------------------------: | :-: | :--: | | Sex | | | | | Male | | 50 | 19.4 | | Monthly spending | | | | |     0–15,000 TL |     0–15,000 TL | 102 | 39.5 | |     15,001–30,000 TL |     15,001–30,000 TL | 88 | 34.1 | |     30,001–45,000 TL |     30,001–45,000 TL | 35 | 13.6 | |     Over 45,000 TL |     All 4 groups over 45,000 TL | 33 | 12.8 | | Highest education | | | | |     Non-university |     High school degree, Associate degree | 181 | 70.2 | |     University |     Bachelor’s degree, Master’s degree, PhD | 77 | 29.8 | **Table 7**\ *Collapsed groups for monthly spending and highest education* _Note_. Frequencies for the original categories are reported in Table 1 (see the Methods section) ### 4.3.1. Overall Empathy Table 8 presents the results of the hierarchical regression analysis predicting overall empathy.

Table 8
Overall empathy hierarchical regression model

Predictor B SE β t p
Step 1: Control Variables
Age -6.60 8.54 -.06 -0.77 .440

Sex

(ref = Male)
    Female 9.43 1.76 .32 5.36 < .001

Highest Education

(ref = Non-university)
    University -0.17 1.82 -.01 -0.09 .925

Monthly Spending

(ref = 0–15,000 TL)
    15,001–30,000 TL 1.74 1.63 .07 1.07 .287
    30,001–45,000 TL 1.00 2.17 .03 0.46 .645
    Above 45,000 TL 5.49 2.28 .16 2.41 .017
Step 2: Main Predictors
Books 2.37 0.96 .15 2.48 .014
ECQ 0.12 0.05 .19 2.39 .018
STAI–S 0.02 0.07 .02 0.33 .742
STAI–T -0.08 0.11 -.06 -0.67 .504
Step 3: Interaction Terms
Books × ECQ -0.05 0.07 -.06 -0.78 .435
Books × STAI–S 0.00 0.00 .01 0.10 .917
Books × STAI–T 0.00 0.01 .01 0.17 .866
_Note_. Step 1: _F_(6, 251) = 6.57, _p_ = \< .001, *R*² = .14. Step 2: Δ*R*² = .05, _F_change(4, 247) = 3.91, _p_change = .004; _F_(10, 247) = 5.69, _p_ \< .001, *R*² = .19. Step 3: Δ*R*² = .00, _F_change(3, 244) = 0.29, _p_change = .831; _F_(13, 244) = 4.41, _p_ \< .001, *R*² = .19. Demographic variables explained 14% of the variance in empathy, *R*² = .14, _F_(6, 251) = 6.57, _p_ \< .001. Being female significantly predicted higher empathy scores (_β_ = .32, _p_ \< .001), and participants with higher income (above 45,000 TL) showed significantly greater empathy (_β_ = .16, _p_ = .017); other covariates were non-significant (\|_β_\| ≤ .07, *p*s ≥ .287). Adding the main predictors accounted for an additional 5% of variance, *ΔR*² = .05, _F_change(4, 247) = 3.91, _p_change = .004; total *R*² = .19, _F_(10, 247) = 5.69, _p_ \< .001. Both book count (_β_ = .15, _p_ = .014) and existential concerns (_β_ = .19, _p_ = .018) were significant positive predictors of empathy, while trait and state anxiety were not (\|_β_\| ≤ .06, *p*s ≥ .504). Interaction terms did not contribute additional explanatory power, *ΔR*² = .00, _F_change(3, 244) = 0.29, _p_change = .831; final *R*² remained at .19, _F_(13, 244) = 4.41, _p_ \< .001. Overall, greater literary exposure and existential concerns were associated with higher empathy, independent of sex and income, while anxiety and interaction effects played no meaningful role. ### 4.3.2. Overall Mentalisation Table 9 presents the results of the hierarchical regression analysis predicting overall mentalisation.

Table 9
Overall mentalisation hierarchical regression model

Predictor B SE β t p
Step 1: Control Variables
Age 23.65 9.33 .19 2.53 .012

Sex

(ref = Male)
    Female 4.48 1.92 .14 2.33 .021

Highest Education

(ref = Non-University)
    University -1.15 1.99 -.04 -0.58 .563

Monthly Spending

(ref = 0–15,000 TL)
    15,001–30,000 TL -1.80 1.78 -.07 -1.01 .312
    30,001–45,000 TL -2.37 2.37 -.07 -1.00 .318
    Above 45,000 TL 0.64 2.49 .02 0.26 .797
Step 2: Main Predictors
Books 1.01 1.01 .06 1.01 .315
ECQ 0.03 0.05 .05 0.57 .567
STAI–S 0.04 0.07 .04 0.48 .632
STAI–T -0.54 0.12 -.40 -4.55 < .001
Step 3: Interaction Terms
Books × ECQ -0.05 0.07 -.06 -0.75 .456
Books × STAI–S -0.01 0.01 -.07 -0.91 .364
Books × STAI–T -0.00 0.00 -.02 -0.22 .827
_Note_. Step 1: _F_(6, 251) = 2.54, _p_ = = .021, *R*² = .06. Step 2: Δ*R*² = .12, _F_change(4, 247) = 9.04, _p_change \< .001; _F_(10, 247) = 5.34, _p_ \< .001, *R*² = .18. Step 3: Δ*R*² = .01, _F_change(3, 244) = 0.64, _p_change = .592; _F_(13, 244) = 4.23, _p_ \< .001, *R*² = .18. Demographic variables accounted for 6% of the variance in mentalisation, *R*² = .06, _F_(6, 251) = 2.54, _p_ = .021. Age significantly predicted higher mentalisation (_β_ = .19, _p_ = .012), and female sex was associated with higher scores (_β_ = .14, _p_ = .021); education and income were not significant predictors (\|_β_\| ≤ .07, *p*s ≥ .312). Adding the main predictors increased explained variance by 12%, *ΔR*² = .12, _F_change(4, 247) = 9.04, _p_change \< .001; total *R*² = .18, _F_(10, 247) = 5.34, _p_ \< .001. Trait anxiety was the only significant predictor in this block (_β_ = –.40, _p_ \< .001); book count, existential concerns, and state anxiety were not significant (\|_β_\| ≤ .06, *p*s ≥ .315). Interaction terms did not contribute additional variance, *ΔR*² = .01, _F_change(3, 244) = 0.64, _p_ = .592; final *R*² remained .18, _F_(13, 244) = 4.23, _p_ \< .001. Overall, greater age, female sex, and lower trait anxiety predicted higher mentalisation, with no evidence that literary exposure or its interactions significantly contributed to the model. ### 4.3.3. Cognitive Empathy Table 10 presents the results of the hierarchical regression analysis predicting cognitive empathy.

Table 10
Cognitive empathy hierarchical regression model

Predictor B SE β t p
Step 1: Control Variables
Age -5.89 5.99 -.07 -0.98 .326

Sex

(ref = Male)
    Female 4.21 1.23 .21 3.42 < .001

Highest Education

(ref = Non-University)
    University 0.21 1.28 .01 0.17 .868

Monthly Spending

(ref = 0–15,000 TL)
    15,001–30,000 TL 1.88 1.14 .11 1.64 .102
    30,001–45,000 TL 0.28 1.52 .01 0.19 .853
    Above 45,000 TL 3.60 1.60 .15 2.25 .025
Step 2: Main Predictors
Books 1.37 0.67 .12 2.04 .043
ECQ 0.09 0.04 .20 2.44 .015
STAI–S 0.01 0.05 .01 0.16 .875
STAI–T -0.16 0.08 -.18 -2.05 .041
Step 3: Interaction Terms
Books × ECQ -0.02 0.05 -.03 -0.40 .688
Books × STAI–S 0.00 0.00 .00 -0.06 .954
Books × STAI–T 0.00 0.00 -.03 -0.40 .690
_Note_. Step 1: _F_(6, 251) = 3.51, _p_ = .002, *R*² = .08. Step 2: Δ*R*² = .04, _F_change(4, 247) = 3.05, _p_change = .018; _F_(10, 247) = 3.39, _p_ \< .001, *R*² = .12. Step 3: Δ*R*² = .00, _F_change(3, 244) = 0.14, _p_change = .935; _F_(13, 244) = 2.62, _p_ = .002, *R*² = .12. Demographic control variables accounted for 8% of the variance in cognitive empathy, *R*² = .08, _F_(6, 251) = 3.51, _p_ = .002. Female sex predicted higher cognitive empathy (_β_ = .21, _p_ \< .001), and higher monthly spending (above 45,000 TL) was also associated with higher scores (_β_ = .15, _p_ = .025); other covariates were non-significant (\|_β_\| ≤ .11, *p*s ≥ .102). The addition of main predictors explained an additional 4% of variance, *ΔR*² = .04, _F_change(4, 247) = 3.05, _p_change = .018; total *R*² = .12, _F_(10, 247) = 3.39, _p_ \< .001. Greater existential concerns (_β_ = .20, _p_ = .015), higher book count (_β_ = .12, _p_ = .043), and lower trait anxiety (_β_ = –.18, _p_ = .041) significantly predicted higher cognitive empathy, while state anxiety was not significant (_β_ = .01, _p_ = .875). Interaction terms did not improve model fit, *ΔR*² = .00, _F_change(3, 244) = 0.14, _p_change = .935; final *R*² = .12, _F_(13, 244) = 2.62, _p_ = .002. No interaction term reached significance (\|_β_\| ≤ .03, *p*s ≥ .688). After accounting for demographic variables, cognitive empathy was significantly predicted by existential concerns, literary exposure, and trait anxiety, while interactions yielded no additional explanatory value. ### 4.3.4. Affective Empathy Table 11 presents the results of the hierarchical regression analysis predicting affective empathy.

Table 11
Affective empathy hierarchical regression model

Predictor B SE β t p
Step 1: Control Variables
Age -0.71 3.87 -.01 -0.18 .854

Sex

(ref = Male)
    Female 5.21 0.80 .38 6.54 < .001

Highest Education

(ref = Non-University)
    University -0.38 0.83 -.03 -0.46 .643

Monthly Spending

(ref = 0–15,000 TL)
    15,000–30,000 TL -0.14 0.74 -.01 -0.19 .850
    30,000–45,000 TL 0.72 0.99 .05 0.73 .466
    Above 45,000 TL 1.89 1.03 .12 1.82 .069
Step 2: Main Predictors
Books 0.99 0.42 .13 2.34 .020
ECQ 0.03 0.02 .11 1.49 .137
STAI–S 0.02 0.03 .03 0.49 .624
STAI–T 0.09 0.05 0.15 1.76 .080
Step 3: Interaction Terms
Books × ECQ -0.03 0.03 -.08 -1.12 .262
Books × STAI–S 0.00 0.00 .02 0.33 .744
Books × STAI–T 0.00 0.00 .07 1.02 .308
_Note_. Step 1: _F_(6, 251) = 8.59, _p_ \< .001, *R*² = .17. Step 2: Δ*R*² = .08, _F_change(4, 247) = 6.95, _p_change \< .001; _F_(10, 247) = 8.42, _p_ \< .001, *R*² = .25. Step 3: Δ*R*² = .01, _F_change(3, 244) = 0.92, _p_change = .430; _F_(13, 244) = 6.69, _p_ \< .001, *R*² = .26. Control variables explained 17% of the variance in affective empathy, *R*² = .17, _F_(6, 251) = 8.59, _p_ \< .001. Being female was a strong positive predictor (_β_ = .38, _p_ \< .001), while age, education, and income brackets were non-significant (\|_β_\| ≤ .12, *p*s ≥ .069). Adding the main predictors increased explained variance by 8%, *ΔR*² = .08, Fchange(4, 247) = 6.95, _p_change \< .001, yielding *R*² = .25, _F_(10, 247) = 8.42, _p_ \< .001. Literary exposure was a significant positive predictor (_β_ = .13, _p_ = .020), and trait anxiety showed a marginal positive association (_β_ = .15, _p_ = .080); existential concerns and state anxiety were non-significant (\|_β_\| ≤ .11, *p*s ≥ .137). Interaction terms added only 1% to the explained variance, *ΔR*² = .01, _F_change(3, 244) = 0.92, _p_change = .430, with no significant interactions (\|_β_\| ≤ .08, *p*s ≥ .262). Overall, affective empathy was higher in females and among those with greater literary exposure, and trait anxiety showed a trend-level positive association, though this did not reach significance. ### 4.3.5. Other-Related Mentalisation Table 12 presents the results of the hierarchical regression analysis predicting other-oriented mentalisation.

Table 12
Other-related mentalisation hierarchical regression model

Predictor B SE β t p
Step 1: Control Variables
Age 2.18 3.80 .04 0.57 .567

Sex

(ref = Male)
    Female 1.85 0.78 .15 2.36 .019

Highest Education

(ref = Non-University)
    University 0.57 0.81 .05 0.70 .486

Monthly Spending

(ref = 0–15,000 TL)
    15,001–30,000 TL -0.51 0.73 -.05 -0.71 .480
    30,001–45,000 TL -0.54 0.97 -.04 -0.56 .579
    Above 45,000 TL 0.34 1.02 .02 0.33 .739
Step 2: Main Predictors
Books 0.48 0.43 .07 1.12 .265
ECQ 0.06 0.02 .23 2.75 .006
STAI–S 0.02 0.03 .06 0.78 .434
STAI–T -0.18 0.05 -.33 -3.61 < .001
Step 3: Interaction Terms
Books × ECQ -0.04 0.03 -.10 -1.30 .196
Books × STAI–S 0.00 0.00 .00 0.04 .972
Books × STAI–T 0.00 0.00 -.05 -0.70 .482
_Note_. Step 1: _F_(6, 251) = 1.46, _p_ = .191, *R*² = .03. Step 2: Δ*R*² = .06, _F_change(4, 247) = 4.12, _p_change = .003; _F_(10, 247) = 2.57, _p_ = .006, *R*² = .09. Step 3: Δ*R*² = .01, _F_change(3, 244) = 0.94, _p_change = .423; _F_(13, 244) = 2.19, _p_ = .001, *R*² = .10. Demographic variables accounted for 3% of the variance in other-related mentalisation, *R*² = .03, _F_(6, 251) = 1.46, _p_ = .191. None of the control variables reached significance (\|_β_\| ≤ .15, *p*s ≥ .019). Adding the main predictors increased explained variance by 6%, *ΔR*² = .06, _F_change(4, 247) = 4.12, _p_change = .003; total *R*² = .09, _F_(10, 247) = 2.57, _p_ = .006. Higher existential concerns significantly predicted greater other-related mentalisation (_β_ = .23, _p_ = .006), while higher trait anxiety was associated with lower scores (_β_ = –.33, _p_ \< .001); book count and state anxiety were non-significant (\|_β_\| ≤ .08, *p*s ≥ .153). Interaction terms added only 1% to the variance, *ΔR*² = .01, _F_change(3, 244) = 0.94, _p_change = .423; final *R*² = .10, _F_(13, 244) = 2.19, _p_ = .001. No interaction reached significance (\|_β_\| ≤ .10, *p*s ≥ .196). Overall, greater existential concerns and lower trait anxiety were associated with stronger other-related mentalisation, while literary exposure and its interactions did not yield meaningful effects. ### 4.3.6. Self-Related Mentalisation Table 13 presents the results of the hierarchical regression analysis predicting self-oriented mentalisation.

Table 13
Self-related mentalisation hierarchical regression model

Predictor B SE β t p
Step 1: Control Variables
Age 15.14 5.17 .22 2.93 .004

Sex

(ref = Male)
    Female -0.37 1.06 -.02 -0.34 .731

Highest Education

(ref = Non-University)
    University -0.87 1.10 -0.06 -0.79 .430

Monthly Spending

(ref = 0–15,000 TL)
    15,000–30,000 TL -0.97 0.99 -.07 -0.98 .329
    30,000–45,000 TL -0.27 1.31 -.01 -0.20 .838
    Above 45,000 TL -0.28 1.38 -.01 -0.21 .837
Step 2: Main Predictors
Books -0.02 0.47 .00 -0.04 .964
ECQ -0.11 0.02 -.30 -4.39 < .001
STAI–S 0.02 0.03 .03 0.47 .637
STAI–T -0.30 0.06 -.40 -5.47 < .001
Step 3: Interaction Terms
Books × ECQ 0.03 0.03 .06 0.92 .358
Books × STAI–S 0.00 0.00 -.08 -1.26 .209
Books × STAI–T 0.00 0.00 -.06 -0.93 .352
_Note_. Step 1: _F_(6, 251) = 1.74, _p_ = .113, *R*² = .04. Step 2: Δ*R*² = .37, _F_change(4, 247) = 38.04, _p_change \< .001; _F_(10, 247) = 16.87, _p_ \< .001, *R*² = .41. Step 3: Δ*R*² = .01, _F_change(3, 244) = 0.89, _p_change = .445; _F_(13, 244) = 13.17, _p_ \< .001, *R*² = .41. Demographic variables accounted for 4% of the variance in self-related mentalisation *R*² = .04, _F_(6, 251) = 1.74, _p_ = .113. Age significantly predicted higher self-related mentalisation (_β_ = .22, _p_ = .004), whereas sex, education, and monthly spending were non-significant (\|_β_\| ≤ .07, *p*s ≥ .329). Adding the main predictors markedly increased explained variance, *ΔR*² = .37, _F_change(4, 247) = 38.04, _p_change \< .001; total *R*² = .41, _F_(10, 247) = 16.87, _p_ \< .001. Greater existential concerns (_β_ = –.30, _p_ \< .001) and higher trait anxiety (_β_ = –.40, _p_ \< .001) predicted lower self-related mentalisation, whereas book count and state anxiety were non-significant (\|_β_\| ≤ .03, *p*s ≥ .637). Interaction terms did not improve the model, *ΔR*² = .01, _F_change(3, 244) = 0.89, _p_change = .445; final *R*² = .41, _F_(13, 244) = 13.17, _p_ \< .001. No interaction reached significance (\|_β_\| ≤ .08, *p*s ≥ .209). Overall, older participants reported higher self-related mentalisation, while existential concerns and trait anxiety were associated with lower scores; literary exposure, state anxiety, and their interactions did not yield reliable effects.