Popular, rejected

2 downloads 0 Views 384KB Size Report
Children came from 25 early childhood classrooms from 12 primary schools located in a range of neighbourhoods ..... Handleiding. Amsterdam, the Netherlands: ...
How to cite this article: van der Wilt, F., van der Veen, C., van Kruistum, C., & van Oers, B. (2018). Popular, rejected, neglected, controversial, or average: Do young children of different sociometric groups differ in their level of oral communicative competence? Social Development, 00, 1-15. https://doi.org/10.1111/sode.12316

1

Abstract Children’s sociometric status refers to their position within the peer group and plays a major role in their future social-cognitive development. It is therefore important to investigate factors that are related to it. Although it has been suggested that one of these factors is children’s level of oral communicative competence, little attention has been paid to its potential role. Therefore, the present study investigated sociometric group differences in the level of oral communicative competence in a sample of N = 570 children in early childhood education. Sociometric status was measured using a nomination procedure. Based on peer nominations, children were categorized into five sociometric groups: (1) popular (generally well-liked), (2) rejected (generally disliked), (3) neglected (low visibility and neither liked nor disliked), (4) controversial (high visibility and both liked and disliked), and (5) average (at or about the mean on both likability and visibility). In addition, children’s level of oral communicative competence was assessed with the Nijmegen Test for Pragmatics. Results of multi-level analyses revealed significant sociometric group differences: Children who were rejected or neglected by their peers exhibited lower levels of oral communicative competence than average children. Based on these findings, early childhood teachers are encouraged to pay more explicit attention to the promotion of their pupils’ oral communicative competence.

2

Introduction Peer relationships have been identified as one of the most important predictors of children’s future social-cognitive functioning (Asher & Coie, 1990). Meaningful contacts with peers provide children opportunities to learn to play with others, control feelings of aggression, and solve problems that may arise in social interactions (Hay, Payne, & Chadwick, 2004; Mostow, Izard, Fine, & Trentacosta, 2002). It is no surprise then that over the past decades substantial research has been conducted into children’s peer experiences (for a review, see Hay et al., 2004). The majority of these studies were aimed at understanding the nature and meaning of problematic peer relationships (e.g., Deater-Deckard, 2001). A key reason for this research interest is the well-established association between peer relationship problems and a range of negative developmental outcomes. For example, previous research has shown that children who have difficulties in building positive relationships with peers are more likely to exhibit disruptive behaviour, to experience feelings of loneliness, and to perform poorly in school (Flook, Repetti, & Ullman, 2005; Pedersen, Vitaro, Barker, & Borge, 2007; Woodhouse, Dykas, & Cassidy, 2012). As the extent to which children experience difficulties with peers partly depends on their position within the peer group (Hay et al., 2004), an important line of research concerns the investigation of factors that are related to children’s sociometric status. To measure a child’s sociometric status, Moreno (1934) developed the sociometric method. The most commonly used approach in sociometric research is the nomination procedure, which requires children to nominate peers whom they like (resulting in positive nominations) and whom they dislike (resulting in negative nominations; Gifford-Smith & Brownell, 2003). At the most basic level, peer acceptance is operationalized as the number of positive nominations a child received whereas the number of negative nominations indicates a child’s level of rejection by peers (Bukowski, Sippola, Hoza, & Newcomb, 2000). Usually, these positive and negative nominations are also combined in order to yield measures of social preference and social impact (Bukowski et al., 2000). Social preference is a measure of relative likability and is operationalized as the number of positive nominations minus the number of negative nominations. Social impact refers to a child’s visibility and is indicated by summing the positive and negative nominations. Based on these four dimensions (i.e., acceptance, rejection, preference, and impact), children can be categorized into different sociometric groups: (1) popular (generally well-liked), (2) rejected (generally disliked), (3) neglected (low visibility and neither liked nor disliked), (4) controversial (high visibility and both liked and disliked), 3

and (5) average (at or about the mean on both likability and visibility; Bukowski et al., 2000; Gifford-Smith & Brownell, 2003). Although the relative number of children in each group varies from study to study, on average, 11% is identified as popular, 13% as rejected, 9% as neglected, 7% as controversial, and 60% as average (Newcomb, Bukowski, & Pattee, 1993; see also Nelson, Burner, Coyne, Hart, & Robinson, 2016). Research has shown that children who are assigned to different sociometric groups differ in their behavioural profiles (e.g., Coie, Dodge, & Coppotelli, 1982; Nelson, Robinson, Hart, Albano, Marshall, 2010). Detailed descriptions of sociometric group differences have been provided in a meta-analysis (Newcomb et al., 1993) and a review by Gifford-Smith and Brownell (2003). Based on their analyses, this paragraph provides an overview of the most important correlates of each sociometric category. In particular, children who are classified as popular have been characterized as having superior social abilities: Compared to average children, popular children show greater social problem-solving skills, are more frequently engaged in pleasurable peer interactions, and have low levels of aggression. On the contrary, rejected children tend to be more aggressive and disruptive, and less sociable than average children. Children classified as neglected have been found to be less aggressive, but also less sociable than average children. In fact, they are engaged in fewer social interactions than average children and are generally more withdrawn. Controversial children have a unique behavioural reputation: They are usually as sociable as popular children, but are also equally aggressive as or even more aggressive than rejected children. Controversial children seem to differ from rejected children in that they can buffer the negative effects of their aggressive behaviour with their advanced social skills. Finally, children considered to be average are generally regarded as a comparison group when assessing the behavioural correlates of the more extreme sociometric groups. The behaviour of average children is described in terms of the extent to which children in the extreme sociometric groups deviate from it. In order to understand not only how but also why children exhibit a particular behavioural profile, it is important to investigate underlying factors (Parker, Rubin, Erath, Wojslawowicz, & Buskirk, 2006). For example, why do some children behave aggressively in the first place? In finding explanations for the way children behave, Vygotsky’s cultural-historical theory, in which language plays a major role, could be particularly helpful (Vygotsky, 1978). Vygotsky regarded the initial function of language to be social, namely to enable children to interact with others and regulate others’ behaviour. Furthermore, he suggested that children subsequently 4

learn to use language to control their own behaviour by planning, coordinating, and reviewing their actions (Vygotsky, 1978). Outcomes of empirical studies provide support for the relation between children’s language competence and their behaviour (e.g., Bonica, Arnold, Fisher, Zeljo, & Yershova, 2013; Gremillion & Martel, 2014; Ostrov & Godleski, 2007; Rescorla, Ross, & McClure, 2007). It has been demonstrated, for example, that preschoolers who behave disruptively also experience language difficulties (Gremillion & Martel, 2014). As sociometric group differences have been found in children’s behaviour, and language competence seems to play a role in regulating behaviour, one might expect that differences in language competence exist between sociometric groups. To examine this hypothesis, the present study specifically investigated sociometric group differences in oral communicative competence. Oral communicative competence is a broad, complex, and multifaceted concept. According to the original theory of Dell Hymes (1972), it includes knowledge of (1) grammar and vocabulary, (2) rules of speaking, and (3) how to use and respond to different types of speech acts. Being a competent communicator requires a speaker to master the structural elements of language, such as the rules of grammar (Hymes, 1972; see also Archer, 2000; Celce-Murcia, 2008). However, in addition to this linguistic competence, one must also be able to use the structural elements of language appropriately in different social situations: “There are rules of use without which the rules of grammar would be useless” (Hymes, 1972, p. 278). Hence, oral communicative competence implies knowledge of syntax, morphology, and phonology, as well as the sociolinguistic competence that allows one to use language in an acceptable and effective manner. As the appropriateness of language use depends on settings, topics, and relationships among people, it is vital to take the social context into account in becoming communicatively competent (Archer, 2000; Celce-Murcia, 2008). One such specific context is children’s peer group. Children who are not able to adapt their language use to this communicative context are likely to encounter misunderstandings and conflicts in peer interactions (Menting, van Lier, & Koot, 2011). Difficulties in interactions with peers might, in turn, increase a child’s risk of developing problematic peer relationships (McCabe & Meller, 2004; Nӕrland, 2011). Conversely, difficulties in peer relationships could shut children out of the type of peer interactions that would support the development of oral communicative competence (Gulay, 2011; Parker et al., 2006). Hence, it seems reasonable to suppose a relation between children’s sociometric status and their level of oral communicative competence.

5

Although scarce, findings of previous research support such a relation. For example, in a smallscale explorative study it was found that, in case of boys, a high level of peer acceptance was associated with a high level of oral communicative competence (van der Wilt, van Kruistum, van der Veen, & van Oers, 2016). In addition, a follow-up on this study demonstrated that children who were frequently rejected by their peers showed low levels of oral communicative competence (van der Wilt, van der Veen, van Kruistum, & van Oers, accepted). However, acceptance and rejection are only two dimensions of sociometric status. In the studies mentioned before, no distinction was made between, for example, children who are rejected by their peers and those who are neglected by them. Although the explorative nature of these previous studies justified the exclusion of other dimensions of sociometric status, more finetuned analyses are required of the relation of sociometric status and oral communicative competence. Unfortunately, previous studies in which sociometric groups were distinguished in the context of communicative skills exclusively concerned small-scale, qualitative studies (e.g., Black & Hazen, 1990). As a result, drawing firm conclusions about the level of oral communicative competence of children assigned to different sociometric groups (i.e., neglected, controversial, etc.) has been impossible thus far. The present study therefore investigated differences in oral communicative competence between all five sociometric groups in a large sample. Including each sociometric status makes it possible to develop a more differentiated understanding of how children’s oral communicative competence might differ according to their sociometric status. In investigating sociometric group differences in oral communicative competence, the present study specifically focused on early childhood. Early childhood is a period in which children’s vocabulary develops rapidly, perspective-taking skills are acquired, and children learn to adapt their speech and communication to the needs of others (Gulay, 2011; Nӕrland, 2011). Furthermore, during this period, children begin to spend an increasing amount of time in a fixed peer group (i.e., their own class), and both the number of peers children come into contact with and the quality of peer interactions change; peers clearly start to form a separate dimension of children’s social life, alongside the family (Gulay, 2011; Nӕrland, 2011). As early childhood is the period in which children’s communicative abilities develop rapidly, and children start to experience the importance of the peer group (Hay et al., 2004), the focus of the present study was on children who were four to six years. In this age group, children’s level of oral communicative competence tends to vary to a great extent (Hay et al., 2004). The heterogeneity in oral communicative competence, combined with the rapid development in this domain, made 6

the period of early childhood a particularly interesting one for investigating sociometric group differences in oral communicative competence. In summary, the main objective of the current study was to investigate sociometric group differences in young children’s oral communicative competence. Based on previous findings, it was expected that rejected children would exhibit poorer oral communicative competence compared to their average peers. Conversely, it was hypothesized that popular children would have higher levels of oral communicative competence than average children. Given the lack of decisive research with neglected and controversial groups, no predictions were made about the level of oral communicative competence of these children. In addition, past research has demonstrated that age (e.g., Nӕrland, 2011), gender (e.g., van der Wilt et al., 2016), and home language (e.g., von Grünigen, Perren, Nägele, & Alsaker, 2010) are related to children’s sociometric status and/or level of oral communicative competence. These three variables were included in the analyses in order to control for their effect. Finally, as previous research showed gender differences in sociometric group composition (e.g., Menting et al., 2011), the possible interaction between sociometric status and gender was taken into account as well. Method Ethical Considerations The present study concerns a sub-study of a larger research project for which ethical approval was obtained from the Scientific and Ethical Review Board of the Faculty of Behavioural and Movement Sciences of the Vrije Universiteit Amsterdam (van der Veen, 2017). Using social media, early childhood teachers were invited to participate in the present study. Teachers who agreed to take part were informed about the purpose of the research and related practical issues. The parents of the children in the participating classes received an information letter which clearly stated that participation of their children was fully voluntary and could be stopped at any moment. Parents could contact their child’s teacher or the principal researcher in case they did not want their child to participate in the study. No parents objected to participation. During the research, all data were carefully sampled and saved anonymously. Only the main researchers of the project had access to the data. Participants

7

The total sample consisted of N = 570 children (299 boys and 271 girls) from 3.99 to 7.08 years (M = 5.11, SD = 0.69). Children came from 25 early childhood classrooms from 12 primary schools located in a range of neighbourhoods (urban, suburban, and rural) in the Netherlands. Dutch primary schools have eight grades, known as groups, which children attend between the ages of four (group 1) to twelve (group 8). The present study included children from the first two groups of primary school, groups 1 and 2. In the Netherlands, groups 1 and 2 are often combined into the same classroom, containing children between the ages of four to six years. Although groups 1 and 2 are part of the Dutch primary school, they are comparable to kindergarten as, for example, play can be characterized as the leading activity. In the present study, class sizes ranged from 13 to 29 children (M = 22.80, SD = 3.94). Additional background information was available for approximately 75% of the total sample. In this subsample, for 61.9% of children Dutch was the main language spoken at home. Furthermore, 57.0% of the children had the Dutch nationality. Other nationalities were Moroccan (5.4%), Turkish (1.9%), Surinam (2.8%), and other western (3.9%) and non-western (4.7%) backgrounds. Parents had low (9.0%), medium (30.3%), and high (60.7%) education levels. Measures Sociometric status. Children’s sociometric status was measured with a nomination procedure (see for example Gifford-Smith & Brownell, 2003). This procedure is frequently used in research into peer relationships and requires children to nominate peers whom they like and whom they dislike. The reliability of the peer nomination procedure has been examined by Wu, Hart, Draper, and Olsen (2001) by calculating test-retest correlations over an 8-week period in a sample of four- to five-year-olds. With a reliability coefficient of .79, the peer nomination procedure proved to be a reliable method for measuring sociometric status in early childhood. In addition, Cillessen, Bukowski, and Haselager (2000) conducted a meta-analysis into the stability of children’s sociometric status. Results showed that about 50% of popular and rejected children retained their status for a longer period of time (i.e., more than three months). Although the stability of the controversial and neglected groups tended to be lower, the concurrent validity of the five sociometric groups is strong: Sociometric groups consistently have been distinguished from one another, confirming that children of different sociometric groups significantly differ in their behaviour (for a review, see Gifford-Smith & Brownell, 2003; see also Cillessen, 2009).

8

Oral communicative competence. The Nijmegen Test for Pragmatics was used to measure children’s level of oral communicative competence (Embrechts, Mugge, & van Bon, 2005). This Dutch test assesses the oral communicative abilities of children aged four to seven years and consists of a scale model of a house with nine associated pictures displaying the rooms in the house. Originally, the Nijmegen Test for Pragmatics consists of three scales: Communicative Functions, Conversation Skills, and Story Building. Because the entire test takes a long time to complete (i.e., 45-60 minutes), only the subscales Communicative Functions and Conversation Skills were administered in the present study. During the administration of the Nijmegen Test for Pragmatics, children are asked to react to a story that is told about the two children who live in the playhouse. For example, one item goes as follows: “Peter rather wants to play with the blocks. But the blocks are high on the cupboard. Peter cannot reach them. What does Peter ask from dad? Dad…?” Children’s answers are scored dichotomously: One point is assigned if the question is answered correctly whereas zero points are assigned in case of an incorrect answer. As for the previous example, “Do you want to grab the blocks for me?” = 1 and “He cannot reach them” = 0. In the present study, a second rater scored five percent of the administered tests independently. With a Cohen’s Kappa of .86, the inter-rater reliability of the Nijmegen Test for Pragmatics was found to be strong (Landis & Koch, 1977). In addition, outcomes of reliability analyses proved that the internal consistency of the items of the Nijmegen Test for Pragmatics is high (Omega = .91, GLB = .95, Cronbach’s alpha = .91). Correlational analyses demonstrated, however, that the overlap between the two subscales (i.e., Communicative Functions and Conversation Skills) was so large that they could not be distinguished meaningfully (r = .79, N = 570, p < .001). As a consequence, no distinction was made between the two scales in the analyses. Regarding construct validity, previous research demonstrated a significant correlation (r = .48) between the scores on the Nijmegen Test for Pragmatics and those on a grammatical language test (Embrechts et al., 2005). This result indicates that performance on the Nijmegen Test for Pragmatics is related to other aspects of language competence (as would be expected), but that it does measure something distinct as well (indicated by the less than perfect correlation). Procedure Sociometric status. Children’s sociometric status was individually assessed by test assistants who were experienced in administrating tests. To familiarize children with the nomination 9

procedure, they first participated in an orientation task that simulated the procedure. This familiarization process was intended to ensure the validity of the children’s responses. During the orientation task, pictures of various types of food were laid out in front of children. They were asked to point to three types of food they liked and three types they disliked. After the orientation task, the test assistant showed children pictures of their classmates. In order to ensure that children attended to each picture, they were asked to name each peer. Next, children nominated (point to) three peers they liked to play with (positive nominations) and three peers they did not like to play with (negative nominations). Finally, children participated in a task in which they selected three types of toys they liked and three types they disliked. This final task was included to reduce the chance that children would discuss their peer nominations with each other as soon as they returned to their classroom. The number of positive (M = 2.80, SD = 2.04) and negative nominations (M = 2.76, SD = 2.26) were used to determine children’s sociometric status. To control for differences in classroom sizes, these positive and negative nominations were standardized within class by converting them to z-scores. For each child, this resulted in a standardized positive nomination score (PN) and a standardized negative nomination score (NN). Subsequently, social preference scores (SP) were calculated by subtracting the standardized negative nomination scores from the standardized positive nomination scores. Furthermore, social impact scores (SI) were calculated by summing the standardized positive and negative nomination scores. These preference and impact scores were again standardized within class. Based on these scores, and following a formula developed by Coie et al. (1982; see also Nelson et al., 2016), children were classified into one of the following sociometric groups: (1) popular, (2) rejected, (3) neglected, (4) controversial, and (5) average. Table 6.1 lists the criteria used to assign children to the five sociometric groups and presents the descriptive statistics for each group. Table 6.1 Criteria and Descriptive Statistics Separated for the Five Sociometric Groups (N = 570) Sociometric groups Popular

Criteria SP > 1, PN > 0, NN < 0

Mean oral

n

n girls

Mean

(%)

(%)

age (SD)

73

35

5.37

(28.02)

1(12.8%)

(47.9%)

(0.63)

1(5.32)

10

communicative competence (SD)

Rejected

SP < -1, PN < 0, NN > 0

Neglected

SI < -1, PN < 0, NN < 0

Controversial

SI > 1, PN > 0, NN > 0

Average

Not in previous categories

Total

78

29

4.87

(20.60)

1(13.7%)

(37.2%)

(0.71)

1(9.13)

77

45

4.89

(21.61)

1(13.5%)

(58.4%)

(0.65)

1(8.03)

50

15

5.37

(26.45)

11(8.8%)

(30.0%)

(0.61)

1(6.96)

292

147

5.12

(25.53)

1(51.2%)

(50.3%)

(0.68)

1(7.55)

570

271

5.11

(24.73)

(100.0%)

(47.5%)

(0.69)

1(7.89)

Note. SP = standardized social preference score; SI = standardized status impact score; PN = standardized positive nomination score; NN= standardized negative nomination score. Oral communicative competence. The administrations of the Nijmegen Test for Pragmatics were individually performed by a trained test assistant in a quiet room near children’s own classroom. Administration of the shortened version of the test took approximately 20 minutes per child. Test administrations were audio taped in order to score children’s answers afterwards. The level of oral communicative competence was calculated by counting the total number of correct answers. This allowed the scores to range from 0 to 37 where a low score indicated a low level of oral communicative competence and a high score indicated a high level. Data Analysis Plan Data were analysed using the Statistical Package for Social Scientists (SPSS, version 23). Complete data were available for the sociometric method. With regard to oral communicative competence, however, 3.3% of the data were missing on the items of the Nijmegen Test for Pragmatics. Missing values were imputed using the commonly used Expectation-Maximization (EM) method in SPSS after finding no statistically reliable deviation from randomness using Little's MCAR test, X2(272) = 362.27, p = .081. The imputed dataset was used in subsequent analyses. A visual inspection of the data indicated that the scores on oral communicative competence were not normally distributed. To further investigate the distribution of these scores, skewness and kurtosis values were obtained (-1.02 and 0.65, respectively) and confirmed that the data did not follow a normal distribution. Although the assumption of 11

normality was not met, it was expected that the violation of this assumption would not cause any major problems: Most of the parametric techniques are reasonably tolerant of violations of this assumption with large sample sizes (e.g., with N > 200; Gravetter & Wallnau, 2004). Children’s scores on oral communicative competence (level 1, N = 570) were nested within classes (level 2, N = 25), nested within schools (level 3, N = 12). In order to correct for the hierarchical structure of the data, multilevel modelling was applied. For this purpose, linear mixed model analyses with maximum likelihood (ML) estimations were carried out following the procedures of Snijders and Bosker (2004). Seven multilevel models were applied in which parameters were added systematically. Model 1 was the basic null (or intercept only) model which only accounted for random error (S2e) and random effects of classes (S2c) and schools (S2s). That is, scores on oral communicative competence were allowed to vary between children, between classes, and between schools. Next, three control variables were added as fixed effects: age (Model 2), gender (Model 3), and home language (Model 4). In Model 5, sociometric status was added as a fixed effect to test whether children differing in sociometric status differed in their scores on oral communicative competence. In Model 6, the interaction between gender and sociometric status was added to investigate gender differences in the sociometric group differences in oral communicative competence. Finally, in Model 7, the sociometric group differences in oral communicative competence were allowed to vary between classes and schools. Models were compared using the log likelihood ratio tests for model improvement (alpha of 0.05) and effect sizes were calculated using the procedures suggested by Tymms (2004).

Results Preliminary Analyses The role of age. The relation between oral communicative competence and age was explored using Pearson product-moment correlation analysis. There was a strong, positive correlation between the two variables, r = .51, n = 570, p < .001. Older children exhibited higher levels of oral communicative competence than younger children. To explore whether children of 12

different sociometric groups differed in their age, a one-way between-groups analysis of variance (ANOVA) was conducted. There was a statistically significant difference in age between the five sociometric groups: F (4, 560) = 8.84, p < .001. Post-hoc comparisons using Tukey HSD tests indicated that the mean age of rejected children was significantly lower than the mean age of popular, controversial, and average children. In addition, neglected children were significantly younger than children in the popular and controversial groups. Finally, age levels of children in the average group were significantly lower than those of children in the popular group. The role of gender. An independent-samples t-test was conducted to compare levels of oral communicative competence for boys and girls. There was a significant difference in scores for boys (M = 23.98, SD = 7.72) and girls (M = 25.55, SD = 8.03; t (568) = -2.38, p = .018, twotailed): Girls outperformed boys on the Nijmegen Test for Pragmatics. The magnitude of the differences in the means (mean difference = -1.57, 95% CI: -2.86 to -0.27) was small (eta squared = .01). In addition, a Chi-square test for independence indicated significant gender differences in children’s sociometric status, X2(4, N = 570) = 14.12, p = .007, Cramer’s V = .16. Post-hoc analyses, using standardized residuals, revealed that the controversial group contained significantly more boys than girls. No differences in gender composition were found in the other four sociometric groups. The role of home language. To compare the level of oral communicative competence between children with Dutch as the dominant language spoken at home and children with a non-Dutch home language, an independent-samples t-test was performed. Results demonstrated that children for whom Dutch was the main language spoken at home scored significantly higher on the Nijmegen Test for Pragmatics (M = 24.56, SD = 7.56) compared to children with a nonDutch home language (M = 20.94, SD = 9.76; t (99.94) = 3.07, p < .001, two-tailed). The magnitude of the differences in the means (mean difference = 3.62, 95% CI: 1.30 to 5.94) was small (eta squared = .02). Furthermore, to investigate sociometric group differences in home language, a Chi-square test for independence was conducted, indicating no significant differences, X2(4, N = 432) = 6.54, p = .162, Cramer’s V = .12. Multilevel Analyses Sociometric group differences in oral communicative competence. The primary analyses for this study concerned sociometric group differences in children’s level of oral 13

communicative competence. Table 6.2 shows the results of the fit and comparison of the planned models. As can be seen, Model 5 (including sociometric status) fitted the data best (Model 5 versus Model 4, X2(4) = 19.33, p < .001), indicating that, even after controlling for age (Model 2), gender (Model 3), and home language (Model 4), there was a significant effect of sociometric status on oral communicative competence. Adding the interaction between gender and sociometric status (Model 6) did not result in a better model fit (Model 6 versus Model 5, X2(4) = 5.72, p = .221), indicating that differences between the five sociometric groups in scores on oral communicative competence were the same for boys and girls. In addition, allowing the relation between sociometric status and oral communicative competence to vary between classes and schools (Model 7) did not improve the fit of the model either (Model 7 versus Model 5, X2(8) = 1, p = .998), indicating that sociometric group differences in oral communicative competence were not dependent on class and school.

Table 6.2 Fit and Comparison of Nested Models (N = 570) Comparison Model

Npars

-2Loglikelihood Models

1

Basic null model

4

3949.27

2

+ age

5

3764.15

14

2 vs 1

ΔX2

Δdf

p

185.12

1