School urbanization

TEPS includes the information of school urbanization which was provided by MOE, and the urbanization includes: “rural”, “sub-urban”, and “urban”. The reference group is “rural”.

CHAPTER 4 RESULTS

Descriptive Results

Table 7 lists sample characteristics for level-2 variables. The gender ratio is approximately 1:1.

Almost 75% of respondents came from families with incomes ranging between $20,000 and $99,999;

43% of parents are high school graduates, with the next largest group (24%) not having graduated from high school. Over 70% of respondents were in public school, over 60% of their schools located in city, and over 60% of respondents were in general program. Furthermore, approximately 9% and 15% of adolescents experienced parental divorce and parental marital conflict, respectively, during childhood.

Regarding family type, the majority of respondents are from two-parent families without frequent severe marital conflict (72.0%), followed by two-parent families with frequent severe marital conflict (19.2%), divorced single-parent families without frequent severe marital conflict (4.4%), and divorced single-parent families with frequent severe marital conflict (4.0%).

Table 7 Sample Characteristics: Level 2 variables (N=3,958) (no weighting)

Variables Frequency Percentage (%)

Gender

Male 1,967 49.7

Female 1,991 50.3

Household income (per month, in NT)

<$19,999 267 6.7

$20,000~$49,999 1,409 35.6

$50,000~$99,999 1,539 38.9

$100,000~$149,999 480 12.1

$150,000~$199,999 128 3.2

>$200,000 103 2.6

missing 32 0.8 Parental education

Less than senior high school 962 24.3

Senior high school graduate 1,709 43.2

Associate junior collage 698 17.6

Bachelor’s degree 429 10.8

Graduate degree 113 2.9

missing 47 1.2

School program type

General type 2,375 60

Comprehensive type 567 14.3

Vocational type 884 22.3

Junior college type 132 3.3

School type

Private 1,183 29.9

Public 2,775 70.1

School urbanization

Rural 125 3.2

Sub-urban 1,415 35.8

Urban 2,418 61.1

Parental divorce

No divorce 3,621 91.5

Divorce

0~6 years 177 4.5

7~12 years 160 4.0

Parental marital conflict

No conflict 3,340 84.4

Conflict

Conflict 0~6 years only 194 4.9

Conflict 7~12 years only 290 7.3

Conflict both 0~6 and 7~12 years 118 3.0

Missing 16 0.4

Family types

Table 8 lists sample characteristics for level-1 variables. Over 86% of students considered themselves “happy” or “very happy” in wave 1, decreasing slightly to 81% in wave 4. Notably, although the percentage of students who were “happy” increased 27% from wave 1 (54.9%) to wave 4 (69.7%), the percentage of students who were “very happy” decreased 63% from wave 1 (31.5%) to wave 4 (11.5%), and the percentage of students who were “unhappy” increased 100% from wave 1 (7.9%) to wave 2 (16.1%).

Only one in five respondents considered themselves to have “never” or only “occasionally”

experienced depressed mood during their adolescence. Moreover, adolescents who considered

themselves to “sometimes” and “always” have depressed mood increased significantly from w1 to w4:

increases were 97% for “do not want to deal with others”, 104% for “feeling upset”, 58% for “wanting to yell or throw things”, and 198% for “feeling lonely”.

Generally, the frequency of paternal and maternal involvement decreased slightly from junior high school to senior high school. Respondents responding “always” or “sometimes” to the item

“discussing future employment opportunities or education” increased 24.8% for paternal involvement (from 41.6% to 51.9%) and decreased 1.9% for maternal involvement (from 70.6% to 69.3%); such responses to the item “listening to your thoughts” decreased 5.4% for paternal involvement (from 24.2%

Two-parent family without conflict 2,850 72

Two-parent family with conflict 759 19.2

Divorced family without pre-divorced conflict

Divorce 0~6 years without pre-divorced conflict 124 3.1 Divorce 7~12 years without pre-divorced conflict 52 1.3 Divorced family with pre-divorced conflict

Divorce 0~6 years with pre-divorced conflict 68 1.7 Divorce 7~12 years with pre-divorced conflict 92 2.3

missing 13 0.3

to 22.9%) and 1.7% for maternal involvement (from 54.0% to 53.1%); such responses to the item

“supervising homework or exams to understand your academic achievements” decreased 4.3% for paternal involvement (from 48.7% to 46.6%) and 1.8% for maternal involvement (from 70.4% to 69.1%); and such responses to the item “participating in school activities” decreased 11.3% for paternal involvement (from 15.0% to 13.3%) and 5.7% for maternal involvement (from 24.7% to 23.3%).

Table 8. Sample Characteristics: Level 1 variables (N=3,958) (Percent/Mean Reporting, no weighting)

survey waves

Depressed mood (%)

(1) Don’t want to deal with others (2) Feeling upset (3) Wanting to yell or throw things (4) Feeling lonely Never Occa-

times Always Missing

W1 69.6 23.0 4.9 2.2 0.3 48.9 35.9 10.5 4.3 0.5 58.3 28.5 8 4.8 0.4 67 23.8 5.6 3.1 0.5 (1) Discussing future employment or

education (2) Listening to your thinking (3) Supervising homework or exams to

understand achievements (4) Participating school activities Never Occa-

times Always Missing

W1, 2 14.8 36.2 28.5 17.6 1.4 48.3 25.8 16 8.2 1.7 12.4 37.3 27.8 20.9 1.6 55.3 27.6 11.1 3.9 2

W3, 4 10.8 31.9 34.3 17.6 5.5 46.1 24.1 15.3 7.6 7 11.4 35.5 26.9 19.7 6.5 51.1 25.9 9.9 3.4 9.6

(continued) Table 8. Sample Characteristics: Level 1 variables (N=3,958) (Percent/Mean Reporting, no weighting)

survey waves

Maternal Involvement (%) (1) Discussing future employment or

education (2) Listening to your thinking (3) Supervising homework or exams to

understand achievements (4) Participating school activities Never Occa-

sionally

Some-

times Always Missing Never Occa- sionally

Some-

times Always Missing Never Occa- sionally

Some-

times Always Missing Never Occa- sionally

Some-

times Always Missing

W1, 2 5.6 23.2 37.8 32.8 0.6 23.5 22 29.6 24.4 0.5 5.9 23.2 33.9 36.5 0.5 41.5 33 16.5 8.2 0.8

W3, 4 4.6 22.2 37.1 32.2 3.9 22.7 21.3 29 24.1 3 5.2 22.7 33.4 35.7 3 39.2 31.5 15.7 7.6 6

survey waves

Happiness (%)

Very unhappy Unhappy Happy Very happy Missing

W1 2.0 7.9 54.9 31.5 3.7

W2 2.9 12.8 64.3 18.8 1.3

W3 1.9 10.7 65.7 19.6 2.1

W4 2.0 16.1 69.7 11.5 0.7

HLM Results

Tables 9 to 12 list results from models estimating the trajectories of adolescent happiness from 7^th to 12^th grade. The unconditional model (Table 9) contains no predictors but time and was used to test Hypothesis 1, which hypothesized that adolescent mental health changes over time; Table 10 introduced variables reflecting the relationship status and tested Hypothesis 2, which hypothesizes that parental divorce negatively affects adolescent mental health; Table 11 introduced variables reflecting the relationship status and tested Hypothesis 3, which hypothesizes that parental marital conflict negatively affects adolescent mental health; Table 12 introduced variables reflecting the relationship status and tested Hypothesis 4, which hypothesizes that the effects of parental divorce partly depend on the degree of parental pre-divorce marital conflict. The control variables and parental involvement are included in table 10 to 12, and test Hypothesis 5, which hypothesizes that parental involvement can affect adolescent mental health, and Hypothesis 6, which hypothesizes that gender, parental education, household income, school program type, school type, and school urbanization can explain the trajectory of adolescent mental health. The Bayesian Information Criterion (BIC; Schwarz, 1978) is used to compare the relative goodness-of-fit of alternative models with different subsets of predictors(Singer & Willett, 2003). The formula for BIC in the HLM is as follows:

BIC = D +qLn(N)

D: Deviance based on FIML (full maximum likelihood) estimation q: number of estimated parameters

N: overall number of observations

1. The Trajectory of Happiness (1) The Unconditional Model

First, an unconditional model with a random intercept and random slope over time was estimated, including the linear and quadratic function of time, which was measured using respondent centering grade. During middle adolescence, happiness trajectories were curvilinear (Table 9). The negative coefficient (-0.13, p<.001)) for the time variable and the positive coefficient (0.01) for the quadratic time measure reveal a linear decrease in happiness over time, which is offset by exponentially increasing rate of decline in later middle adolescence.

Table 9. The Unconditional Model—Happiness

Effects Unconditional model

Fixed effects

Model for the intercepts (𝛽₀)

Intercept 3.21***

(.02) Model for Time slope (𝛽₁)

Intercept -.13***

(.03) Model for Time² slope (𝛽₂)

Intercept .01

(.01) Random effects

τ₀₀ .17509***

τ₁₁ .08494***

τ₂₂ .00487*

Deviance 28204.46

BIC 28234.11

No. of parameter 10

Note: Total observations in level 1are 15,240; total observations in level 2 are 3,886.

* p < .05. ** p < .01. *** p <.001.

Figure 3 shows the trajectory of adolescent happiness from 7^th grade to 12^th grade. The dotted line represents the level of happiness when the equation only includes time, and the solid line represents

the level of happiness when the equation includes both time and control variables (including: gender, household income, parental education, school program type, school type, and school urbanization). The vertical axle ranged from 1.5 to 3.5 (1 to 4 originally) in all trajectory figures (Figs. 3 to 15) to enable easier distinguishing of the different lines. Both lines indicate that the trajectory of happiness in adolescents changed over time, supporting hypothesis 1-1.

Figure 3. The trajectory of adolescent happiness

(2)The Effects of Parental Divorce

Table 10 lists the effects of parental divorce on adolescent happiness. In model 1, adolescents who experienced parental divorce during the ages of 0~12 years old were recoded using the code 1, and the reference group comprised adolescents without such experience of parental divorce. The result shows that parental divorce during childhood initially reduces adolescent happiness (-.14, p<.05). This result is consistent with previous studies suggesting that parental divorce negatively affects youth outcomes (Kalter & Rembar, 1981; Krein & Beller, 1988; Palosaari & Aro, 1994; Zill, et al., 1993).

Hypothesis 2-1 thus is supported. However, neither the linear coefficients (-.17) nor the quadratic

coefficients (.07) are statistically significant, implying that parental divorce during childhood does not affect the increase in adolescent happiness.

Model 2 examines whether parental divorce occurring in early childhood (0~6 years old) or late childhood (7~12 years old) is more damaging to adolescent happiness compared with the situation of two parent families. The analytical result shows that adolescents whose parents divorced when they were 7~12 years old initially have lower happiness than those from two-parent families (-.25, p<.01), but neither the linear (-.14) nor the quadratic (.08) coefficient are statistically significant, implying the rate of growth of happiness does not differ relative to two parent families. These results surprisingly do not support those of Kerin and Beller (1988), and Zill et al. (1993) which suggested that parental divorce at an early age (0~5.5 years old, and 0~6 years old, respectively) is more harmful for children, but their outcomes are educational attainment rather than mental health.

Table 10. HLM Result: The Effect of Parental Divorce on Adolescent Happiness

Effects

Parental divorce (=1) -.14*

(.07) Age at parental divorce

No divorce (=0) ---

Senior high school graduate .02 Household income per month

<NT$19,999 (=0) --- --- School urbanization

Rural (=0) --- ---

Parental divorce (=1) -.17

(.11) Age at parental divorce

No divorce (=0) --- Household income per month

<NT$19,999 (=0) --- --- School urbanization

Rural (=0) --- ---

(.06) (.06)

Parental divorce (=1) .07

(.04) Age at parental divorce

No divorce (=0) --- Household income per month

<NT$19,999 (=0) --- --- School urbanization

Rural (=0) --- ---

Note: Total observations in level 1are 12,918 (model 1~model 2); total observations in level 2 are 3,748 (model 1~model 2). * p < .05. ** p < .01. *** p <.001. Standard error is in (parentheses).

Figure 4 shows the trajectory of happiness of adolescents from divorce and no-divorce families.

The figure shows that adolescents from divorced families have lower happiness than those from two parent families in 7^th grade, but the rate of decrease does not differ statistically significantly between the two groups. The small solid triangles above the notation in all trajectory figures represent the statistical significance relative to the reference group (the solid blue line), and the hollow triangles represent that the coefficient is not statistically significant, but does closely approach statistical significance relative to with the reference group.

(.06) (.06) Model for Paternal Involvement

Intercept .01(.01) .01

(.01) Model for Maternal Involvement

Intercept .02***

(.00)

.02***

(.00) Random effects

τ₀₀ .16992*** .16874***

τ₁₁ .11342* .11374*

τ₂₂ .00683 .00686

Deviance 23340.74 23329.87

BIC 23937.12 23954.65

No. of parameter 63 66

Figure 4. The trajectory of adolescent happiness – by parental divorce (two groups)

Figure 5 illustrates the trajectory of happiness of adolescents from two parent families, families where the parents divorced when the children were aged 0~6 years old, and families where the parents divorced when the children were aged 7~12 years old. The figure shows that adolescents from families in which parents divorced when their children were in late childhood (7~12 years old) have lower levels of happiness than adolescents in two parent families in 7^th grade, but the growth rate does not differ.

Figure 5. The trajectory of adolescent happiness – by parental divorce (three groups)

(3)The effects of parental marital conflict

Table 11 lists the effects of parental marital conflict on adolescent happiness. In model 1, adolescents who experienced parental marital conflict during 0~12 years old were recorded using the code 1, while the reference group comprised adolescents who had not experienced this. The analytical result shows that parental marital conflict during childhood reduces the adolescent happiness initially (-.15, p<.01), supporting previous studies (Lindsey, et al., 2009; Mechanic & Hansell, 1989). Hypothesis 3-1 thus is supported. However, neither the linear (-.03) nor quadratic coefficients (.03) are statistically significant, implying that parental marital conflict during childhood does not affect the increase in adolescent happiness.

Model 2 examines whether parental conflict occurring only during early childhood (0~6 years old), during late childhood (7~12 years old), or during the whole of childhood (both age 0~6 years and 7~12 years) in families that have undergone divorce is more damaging for adolescent happiness compared with that occurring in two parent families. The comparison results show that adolescents whose parents display conflict in late childhood (7~12 years old) have lower levels of happiness than

those of two-parent families (-.18, p<.01) in the initial level, but neither the linear (-.04) nor quadratic (.04) coefficients are statistically significant, implying that the growth rate of happiness does not differ from that in two parent families.

Table 11. HLM Result: The Effect of Parental Marital Conflict on Adolescent Happiness

Effects Parental marital conflict (=1) -.15**

(.05) Age at parental marital conflict

No conflict (=0) --- Household income per month

<NT$19,999 (=0) --- ---

NT$20,000~NT$49,999 -.21

(.14)

-.21 (.14)

NT$50,000~NT$99,999 -.13 School urbanization

Rural (=0) --- --- Parental marital conflict (=1) -.03

(.07) Age at parental marital conflict

No conflict (=0) ---

(.07) (.07) Household income per month

<NT$19,999 (=0) --- --- School urbanization

Rural (=0) --- --- Parental marital conflict (=1) .03

(.03) Age at parental marital conflict

No conflict (=0) ---

Male .01 Household income per month

<NT$19,999 (=0) --- --- School urbanization

Rural (=0) --- --- Model for Paternal Involvement

Intercept .01

(.01)

.01 (.01) Model for Maternal Involvement

Intercept .02***

(.01)

.02***

(.01) Random effects

Note: Total observations in level 1are 12,918 (model 1~model 2); total observations in level 2 are 3,748 (model 1~model2). * p < .05. ** p < .01. *** p <.001. Standard error is in (parentheses).

Figure 6 shows the trajectory of happiness for adolescents from parental conflict and no parental conflict families. The Figure shows that adolescents from conflict families have lower happiness than those from two parent families in 7^th grade, but the rate of decrease is also quite similar between the two groups.

Figure 6. The trajectory of adolescent happiness – by parental marital conflict (two groups)

Figure 7 shows the trajectory of happiness of adolescents from no parental conflict families, families with conflict occurring when the children are 0~6 years old, families with conflict occurring at

τ₀₀ .16820*** .16805***

τ₁₁ .11271* .11235*

τ₂₂ .00687 .00683

Deviance 23351.04 23346.41

BIC 23947.42 23999.59

No. of parameter 63 69

7~12 years old, and families with conflict occurring at both 0~6 years old and 7~12 years old. The figure shows that adolescents of families with parental conflict when children are 6~12 years old have lower happiness than those of two parent family in 7^th grade, but the rate of decrease is not significant different.

Figure 7. The trajectory of adolescent happiness – by parental marital conflict (four groups)

(4)The effects of family types

Table 12 lists the effects of family types on adolescent happiness by combining the items of

“parental divorce” and “parental marital conflict”. Model 1 creates four family types: two parent family without conflict, two parent family with conflict, divorced family (with divorce occurring when the child was 0~12 years old) without pre-divorce marital conflict, and divorced family (with divorce occurring when the child was 0~12 years old) with pre-divorced marital conflict. The result shows that adolescents from divorced families with pre-divorce marital conflict have the lowest happiness (-.32, p<.001) among all family types in terms of initial level, and the negative linear coefficient (-.26, p<.05) indicates that rate of decrease of happiness for adolescents from divorced families with pre-divorce marital conflict is faster than for two-parent families. Hypothesis 4-1 thus is not supported. This unexpected result does not

correspond with previous studies, suggesting that parental divorce is more harmful for offspring in the absence of parental marital conflict (Amato, et al., 1995; Hanson, 1999). Adolescents from two parent families with marital conflict exhibit the second lowest levels of happiness (-.18, p<.001), supporting previous studies finding that long-term pressure and hostility between parents harms offspring.

Six family types were created in model 2: two parent family without conflict, two parent family with conflict, divorced family (with divorce occurring when the child was 0~6 years old) without pre-divorce marital conflict, divorced family (with divorce occurring when the child was 0~6 years old) with pre-divorce marital conflict, divorced family (with divorce occurring when the child was 7~12 years old) without pre-divorce conflict, and divorced family (with divorce occurring when the child was 7~12 years old) with pre-divorce conflict. Model 2 separately examines the effects of pre-divorce conflict on early childhood (0~6 years old), and late childhood (7~12 years old). The results show that adolescents from families that have undergone divorce (with divorce occurring when the child was 7~12 years old) with pre-divorce conflict have the lowest level of happiness among all family types initially (-.42, p<.001), but do not differ statistically in linear (-.07) or quadratic growth rate (.07).

Table 12. HLM Result: The Effect of Family Types on Adolescent Happiness

Effects Family type (overall age 0~12 years)

Two-parent without conflict (=0) --- Divorced 0~12 years with predivorced

conflict

-.32***

(.10)

Family type (divided to age 0~6 and Household income per month

<NT$19,999 (=0) --- ---

Junior college program -.09 School urbanization

Rural (=0) --- --- Family type (overall age 0~12 years)

Two-parent without conflict (=0) --- Divorced 0~12 years with pre-divorced

conflict

-.07 (.19) Family type (divided to age 0~6 and

7~12 years)

Graduate degree .04 (.15)

.04 (.15) Household income per month

<NT$19,999 (=0) --- --- School urbanization

Rural (=0) --- --- Family type (overall age 0~12 years)

Two-parent without conflict (=0) --- Divorced 0~12 years with pre-divorced

conflict

.06 (.07) Family type (divided to age 0~6 and

7~12 years)

conflict (.10) Household income per month

<NT$19,999 (=0) --- --- School urbanization

Rural (=0) --- --- Model for Paternal Involvement

Intercept .01

(.01)

.01 (.01)

Note: Total observations in level 1are 12,918 (model 1~model 2); total observations in level 2 are 3,748 (model 1~model2). * p < .05. ** p < .01. *** p <.001. Standard error is in (parentheses).

Figure 8 illustrates the trajectory of happiness for adolescents from the four family types. The figure shows that adolescents from divorced families with pre-divorce conflict have lower happiness than those from two parent families in 7^th grade, but the rate of decrease is also quite similar for the two groups.

Figure 8. The trajectory of adolescent happiness – by four family types Model for Maternal Involvement

Intercept .02***

(.01)

.02***

(.00) Random effects

τ₀₀ .16375*** .16300***

τ₁₁ .11244* .11348*

τ₂₂ .00682 .00688

Deviance 23274.86 23258.57

BIC 23928.04 23968.55

No. of parameter 69 75

Figure 9 shows the trajectory of happiness of adolescents from six family types: including two parent families without marital conflict, two parent families with marital conflict, families with parental divorce occurring when the child was 0~6 years old without pre-divorce conflict, families with parental divorce occurring when the child was 0~6 years old with pre-divorce conflict, families with divorce occurring when the child was 7~12 years old without pre-divorced conflict, and families with divorce occurring when the child was 7~12 years old with pre-divorce conflict. The figure shows that

adolescents from two parent families with marital conflict and divorce occurring when the child was 7~12 years old with pre-divorce conflict have significantly lower happiness in grade 7.

Figure 9. The trajectory of adolescent happiness – by six family types

Male adolescents have higher happiness at the initial point (.09, p<.05 in Table 10 to 12) than female adolescents, but do not differ in growth rate. On the other hand, school program types affect the growth rate of happiness but do not differ at the initial level. In general, the linear growth rate of

happiness of adolescents in comprehensive program increases slightly more rapid (.20, p<.05 in table 10

~ table 12), and the quadratic growth rate decreases slightly less rapid (-.07, p<.05 in table 10 ~ table 12) than those in general program. Despite the coefficient being minor, maternal involvement is positively associated with adolescent happiness (.02, p<.001 in Table 10~12). Household income, parental education, school type, and school urbanization do not explain the trajectory of happiness for adolescents. Hypothesis 5 thus is only partially supported. The τ₀₀ of the unconditional model is 0.17509, and therefore the within individual variance changed overtime for happiness and explained all predictors (including parental divorce, marital conflict, parental involvement and control variables) which ranged from 3.0% [(0.17509-0.16992)/0.17509] to 6.9% [(0.17509-0.16300)/0.17509].

2. The Trajectory of Depressed Mood (1) The Unconditional Model

First, an unconditional model with a random intercept and random slope over time was

estimated, including the linear and quadratic functions of time, which were measured uing the centering grade of adolescent. Adolescent depressed mood displayed an opposite trajectory to happiness (Table 13). The positive coefficients (2.18, p<.001) for the time variable and negative coefficients (-0.58, p<.001) for the quadratic time measure indicate that depression increased over time but the rate of

increase slowed.

Table 13. The Unconditional Model—Depressed Mood

Effects Unconditional model

Fixed effects

Model for the intercepts (𝛽₀)

Intercept 6.18***

(.08) Model for Time slope (𝛽1)

Intercept 2.18***

(.12) Model for Time² slope (𝛽2)

Intercept -.58***

(.04) Random effects

τ₀₀ 1.46486***

τ₁₁ 1.05425

τ₂₂ .08024

Deviance 70400.89

BIC 70497.30

No. of parameter 10

Note: Total observations in level 1are 15,380; total observations in level 2 are 3,886.

* p < .05. ** p < .01. *** p <.001. Standard error is in (parentheses).

Figure 10 shows the trajectory of adolescent depressed mood from 7^th grade to 12^th grade. The dotted line represents depressed mood when the equation includes only time, and the solid line

represents the levels of happiness when the equation includes both time and control variables (including gender, parental education, household income, school program type, school type, and school

urbanization). Both lines indicate that the trajectory of adolescent happiness changed over time, and thus hypothesis 1-2 is supported.

Figure 10. The trajectory of adolescent depressed mood

(2)The Effects of Parental Divorce

Table 14 lists the effects of parental divorce on adolescent depressed mood. In model 1,

adolescents who had experienced parental divorce at 0~12 years old were recorded using the code 1, and the reference group comprised those who had not experienced this. The analytical result shows that parental divorce during childhood increases adolescent depressed mood in initial level (.78, p<.001).

This result is consistent with previous studies suggesting that parental divorce negatively affects youth outcome (Kalter & Rembar, 1981; Krein & Beller, 1988; Palosaari & Aro, 1994; Zill, et al., 1993).

Hypothesis 2-2 thus is supported. However, neither the linear coefficient (.82) nor the quadratic coefficient (-.30) statistically significant, implying that parental divorce during childhood does not influence growth of adolescent depressed mood.

Model 2 examines whether parental divorce occurring in early childhood (0~6 years old) or late childhood (7~12 years old) is the more damaging for adolescent depressed mood compared with adolescents from two parent families. The result shows that adolescents whose parents divorced when

Model 2 examines whether parental divorce occurring in early childhood (0~6 years old) or late childhood (7~12 years old) is the more damaging for adolescent depressed mood compared with adolescents from two parent families. The result shows that adolescents whose parents divorced when

在文檔中 The trajectory of adolescent mental health: The effects of parental divorce and marital conflict during childhood (頁 55-111)