Sample
The present study adopted a dual-frame sampling method meaning that samples are drawn independently from two overlapping frames covering the population of interest (Lohr & Rao, 2000). A cluster sampling procedure (Teddlie & Yu, 2007) included children according to age, gender, degree of exposure to military violence, geographical distribution in Palestine. The first sample (A) was composed of 1,166 Palestinian children aged 9 to 14 years (m = 11.58, sd = 1.54). Children were recruited in different primary and lower secondary school locations covering three governorates in West Bank (Nablus, Jericho, and Bethlehem) and Jerusalem. Approximately 50% of participants were boys (n = 586). Children were attending the following school grades: 5th grade (n = 120 and 10.3%), 6th grade (n = 205 and 17.6%), 7th grade (n = 244 and 20.9%), 8th grade (n = 213 and 18.3%), 9th (n = 235 and 20.2%) and 10th (n = 149 and 12.8%). The second sample (B) was composed of 251 Palestinian children aged between 9 and 14 years (m = 11.82, sd = 1.53) recruited in primary and lower secondary schools located in the Gaza Strip. Similarly to sample A, approximately 50% of participants were boys (n = 123). Children were attending the following school grades: 5th grade (n = 16 and 6.4%), 6th grade (n = 40 and 15.9%), 7th grade (n = 54 and 21.5%), 8th grade (n = 50 and 19.9%), 9th (n = 43 and 17.1%) and 10th (n = 48 and 19.1%). The inclusion criteria for the present study (both samples) were being aged between 9 and 14 years at the study time and being Palestinian.
Background
The Palestinian population has been living under military occupation since the 1948 war, which worsened in '67; daily life has been characterized by ongoing experiences of extreme violence, trauma, and severe human rights violations (Human Rights Watch, 2019). The situation has only deteriorated over time. The barrier built by Israel in the West Bank isolates towns and villages, severely restricting Palestinians' freedom of movement and access to rights such as work, education and health (Amnesty International, 2020) and increasing inaccessibility to treatment in a pandemic situation such as covid-19. On the other hand, Gaza is enclosed for the 14th year in a strip in an illegal air, land and maritime blockade (Amnesty International, 2020). The entry of goods and people is controlled and restricted, with a devastating impact on the lives of 2 million inhabitants in Gaza, already severely stretched by the many military attacks over the years. In addition, Israel has continued to impose 'institutionalized discrimination, forcing the displacement of hundreds of Palestinians in both the West Bank and East Jerusalem, following the demolition of homes and other coercive measures (Amnesty International, 2020). As a consequence of these dispossession policies, the number of refugees represents 73.1% of the population of the Gaza Strip and 30.2% of the West Bank (PCBs, 2016).
The situation throughout the Palestinian territories has been dramatically exacerbated by the spread of the Covid-19 pandemic, severely affecting both the West Bank and the Gaza Strip, where the possibility of implementing social distancing and essential hygiene measures is severely compromised the conflict (OECD, 2020). Indeed, in Gaza, health systems already damaged by the blockade and bombing lack medicines, equipment, and professional staff, making it impossible to contain the virus and raising Covid mortality rates. To date (June 2021), there are over 330,000 registered cases of Covid-19 in the Palestinian territories, with 3,777 deaths between the West Bank and Gaza Strip (WHO, 2021).
Procedure
Before taking part in the study, written parental consent was requested and obtained from all children and their families. Participants and their family were both openly informed about the aim and the objectives of the study. All the data were collected anonymously, and the information was analyzed in an aggregate manner. Participation in the study was free, and no monetary or other incentives were offered. The research protocol was administered in the classroom during school hours by local psychologists and researchers trained in the social sciences. Participants were free to withdraw from the study at any time. Completion of the research protocol took 40/50 min. The settings where it was administered were designed to offer a relational space (Hydén, 2014) in which participants would feel free to express their perceptions and opinions (Veronese, Pepe, & Afana, 2016). All phases of the research were conducted following the ethical guidelines of the American Psychological Association (American Psychological Association, 2010) and the principles enshrined in the Declaration of Helsinki (World Medical Association, 2013). None of the authors reported any financial or other conflicts of interest with the research aims and outcomes. Ethical approval was obtained from the Institutional Review Board of the University of Milano-Bicocca (Protocol N. 368)
Questionnaire And Item Development
In terms of questionnaire development, eliciting children's voices (e.g. understanding how they live their lives, how they perceived their living spaces or acting as social actors from a bottom-up perspective) was central in exploring how they deal with their living circumstances. As Lieten (2008) stated, "close reading is needed with a focus on the daily lives of children, their cognitive understanding and emotional reactions, their interactions with each other and with adults of various kinds, and their strategies of action" (p. xvi). Many textbooks reported about children' agency predominantly from a Western-oriented position [see for instance Corsaro (2009) or Qvortrup (2005)], consequently, to understand the cultural-specific bases of agency in Palestinian children, the first step was necessarily to interrogate them through an in-depth semi-structured interview (Stage 1). To this end, interviews were conducted with 75 children (7–13 years old, m = 10.27,sd = 1.38, 68% female) coming from different contexts in the West Bank and Gaza (i.e., rural areas, urban areas, refugee camps), all attending primary school. Interviews aimed at observing and investigating children's everyday agentic practices (Payne, 2011). Through a highly participatory methodology (interactive maps, walk-along interviews), children were asked to draw and then describe a map of their everyday places, both safe and unsafe, to understand better how the places represented were either enhancing or suppressing their agency. Upon completion of the drawing task, 30 children (40%) were asked to guide the researchers inside the represented places to continue the interview, allowing an in-depth investigation of children's agentic practices in their daily life. The areas of inquiry investigated included: descriptive information (e.g., describe the place); emotions (e.g., how do you feel in this place?); behavioural information (e.g., how do you spend your time here), and relational information (e.g., do you come to this place alone?) (for a full description, see Cavazzoni, Fiorini, Sousa, Veronese, 2021).
After conducting the interviews with children, the next step was to translate all recorded materials into English (Stage 1b; see Fig. 1 for the complete process). In translating, we followed the standard procedures of forward- and back-translation (Brislin, 1970). As a result, two practitioners with a background in psychological research methods provided an initial conceptual translation of children's words. They were fluent in English and familiar with English-speaking cultures; at the same time, they were native Arabic speakers in the target Palestinian culture. The research output of this phase was having a complete set of transcribed interviews translated in English in order to analyze the material for 1) identifying main dimensions of war children agency (i.e., conceptualizing the construct of agency from children voices) and 2) defining empirical indicators for saturating each dimension (i.e., operationalizing the constructs by developing a set of items).
Next, four collaborative discussion sessions with eight experts from the field, clinical psychologist, key informant, and PhD students were organized. In the first two sessions (Stage 2a), eight domains of children agency were mapped (social aspects, education, safety and security, freedom of movement, religion and spirituality, national identity, play and leisure, living and political environment) by using bottom-up thematic analysis (Boyatzis, 1998) to analyze interviews. The other two sessions (Stage 2b) were planned to develop approximately items covering different facets of each dimension. For instance, for the domain "freedom of movement", experts worded items such as "I feel free when I am able to reach my destination by myself" or "I can go to places on my own". At the end of this stage of the study, a set of 109 items was finally developed and screened by a panel of three Arab-native speaking and Palestinian academics (Stage 3). Each external reviewer was asked to rate from 1 (not appropriate) to 5 (totally appropriate) the degree to which the item was pertinent in measuring agency in such a cultural context. Cultural salience underpins the formation of semantic domains, defined as "organized set[s] of words, concepts, or sentences, that jointly refer to a single conceptual sphere" (Weller & Romney, 1988, p. 9). In addition, each item was also commented on in term of cultural appropriateness, quality of wording and redundancy. After receiving all the reviewers' assessment, the only item reporting a minimum mean score or four or more were retained. As a result, a set of 42 items covering the initially identified eight dimensions was retained. During stage 4, 228 Palestinian children were involved in the on-site pilot study to evaluate preliminary statistical proprieties (e.g., distribution of scores, checking for potential floor or ceiling effects, missing rates) of the WCAAS-Pal (e.g., on the pool of 42 items). The sample was composed of 54% of boys (n = 123) aged between 9 and 14. The data-gathering phase followed the criteria described in the procedure section of the present article. Analysis of scores revealed items with highly skewed distribution (i.e., mainly ceiling effect, meaning that a high rate of children has maximum scores on the observed variable). Items excluded were: "I want to learn in order to have a job and enhance my future opportunities", "I want to go to school to be successful in my life", "I choose to go to school because it is my right to obtain an education" and "When something bad happens, I prefer to stay close to my family".
Measure
War Child Agency Assessment Scale - Palestinian version (WCAAS-Pal) the questionnaire was developed to evaluate agency in children from a multidimensional perspective. The score of the questionnaire reflected different domains of children agency as emerged from a bottom-up exploratory multi-method approach. The scale consists of items to be rated on a Likert-type scale ranging from 1 (not at all) to 5 (very much). Sample items include "I can identify places that I enjoy and that are relaxing", "I prefer to avoid places that remind me of the martyrs", and "Discovering things about the world helps me feel in control of my life".
Children's Hope Scale (CHS; Snyder et al., 1996): The Children Hope Scale was a quantitative measure assuming that children are goal-oriented, and it was composed of two domains: agency and pathways. Agency thoughts reflect the perception that children can initiate and sustain action toward a desired goal; pathways reflect the children's perceived capability to produce routes to those goals (Snyder et al., 1996). Hope reflects the combination of agentic and pathways thinking toward goals. As a result, a six-item dispositional self-report index was validated for use with children ages 8–16. The scale consists of items to be rated on a Likert-type scale ranging
from 1 (none of the time) to 6 (all the time). Sample items included: "I think I am doing pretty well", "When I have a problem, I can come up with lots of ways to solve it", and "I am doing just as well as other kids my age". Reliability score (Cronbach's alpha, Cronbach, 1951) for the two domains were: agency (α = .687) and pathway (α = .680).
Children Revised Impact of Event Scale-Arabic Version (CRIES-13A; Veronese & Pepe, 2021): The Children Revised Impact of Event Scale-Arabic Version was an adapted version of the Impact of Event Scale (Dyregrov et al., 1996) for the evaluation of traumatic psychological responses in children who have been exposed to ongoing traumatic events in contexts of violence. The original measurement model comprised three dimensions: intrusion, avoidance, and hyperarousal. Items are rated on a four-point scale (0 = not at all, 5 = often). Sample items included: "Do you think about erasing the event that shocked you from your memory?", "Do you have difficulty concentrating?" and "Do you try to avoid thinking about the shocking event?". Reliability score (Cronbach's alpha, Cronbach, 1951) for the three domains were: intrusion (α = .738), avoidance (α = .715) and hyperarousal (α = .732).
The Strategy Of Data Analysis
The exploration of the measurement model of the War Child Agency Assessment Scale - Palestinian version (WCAAS-Pal) followed standard procedures for validating quantitative self-reports (Matsunaga, 2010; Mertler & Reinhart, 2016). This means that we evaluated the model of measurement of WCAAS-Pal using a sequential exploratory factor analysis (EFA) and confirmatory factor analysis (CFA). The results of the EFA suggested a "baseline" factor structure to be further assessed via CFA. Given the ordinal nature of the scores (Likert-type response scale), EFA was conducted on the polychoric matrix (Holgado-Tello et al., 2010) of correlations computed using Lorenzo-Seva and Ferrando's (2015) syntax for SPSS. We also conducted the Kaiser-Meyer-Olkin (KMO) test and Bartlett's test of sphericity (Field, 2009). Parallel analysis eigenvalues were estimated using Monte Carlo PCA for Parallel Analysis 2.3 (Watkins, 2008). Velicer's minimum average partials (MAPs) were computed using O'Connor's (2000) program developed ad hoc for SPSS. The EFA comprised estimation via principal axis factoring (Briggs & MacCallum, 2003) and oblimin rotation (Darton, 1980). We applied Kaiser's criterion (K1; Kaiser, 1960) to identify the most appropriate number of factors to be retained and parallel analysis (Horn, 1965) to further inform our decisions about the most appropriate factor structure for our adapted version of the questionnaire. Parallel analysis (PA) is a data simulation technique that compares the eigenvalues of the factors extracted from the observed data with random data matrices of comparable size (Hayton et al., 2004). We also estimated Velicer's MAPs (Velicer et al., 2000). Caron (2018) observed that when the component structure is moderately oblique, MAP offers a more accurate indication of the most appropriate number of factors than does PA. Therefore, we only adopted factor loadings (λ) greater than .40 (Hair et al., 2006), dropping items that loaded on more than one factor (Costello & Osborne, 2005).
The measurement model identified from the EFA was next explored to CFA to evaluate further evidence of construct validity (Gagne & Hancock, 2006). CFA was a statistical technique that supported the development of quantitative tools by testing a given measurement model's degree of fit with empirical data (Kline, 2010). In the present paper, the following absolute and relative fit indexes were adopted: χ2, normed chi-square (NC), root mean square error of approximation (RMSEA), standardized root mean square residual (SRMR), Nor-Normative Fit Index (NNFI) and comparative fit index (CFI). Model fit was taken to be accepted if NC was < 3.0, RMSEA < .08, SRMR < .08, NNFI and CFI > .95 (Morin et al., 2013). Finally, the Akaike information criterion (AIC) was adopted to compare models. First, as recommended by Hettie (1985), a unidimensional solution was estimated (with all indicators loading on a single factor, Model 1), and then the baseline model (Model 2) was specified (seven latent variables and 22 empirical indicators).
The subsequent stage of statistical analysis involved using multiple-group CFA (MGCFA; Byrne, 1998) to test the factor structure of the questionnaire in gender-based cohorts of participants to enhance the ecological validity of the results (Cavioni et al. 2020). Measurement invariance between boys and girls was to be supported when configural invariance (when the same items are associated with the same factors in each group), metric invariance (when factor loadings are set to be equal across groups), scalar invariance (when factor loadings and item intercepts are constrained to be equal across groups), and full invariance (when factor loadings, item intercepts, and residuals are constrained to be equal across groups) were accepted. We set the cutoff criteria for rejecting invariance at Δ > .01 for both ΔRMSEA and ΔSRMR (Chen, 2007) and a chi-square difference (Δχ2) that was statistically significant at the p < .01 level (Milfont & Fischer, 2010). The different types of invariance are hierarchically ordered, meaning that the MGCFA procedure ends at the lowest level of invariance that fails to be satisfied (for further details, see Cheung & Rensvold, 2002). We also estimated Mahalanobis' distance (p < 0.001) to detect potentially multivariate outliers; no cases needed to be removed from the dataset. Analysis of missing value revealed a mean score of 1% for blank data; consequently, missing values were replaced at random (Donders et al., 2006). Statistical Package for the Social Sciences 25.0 (SPSS; Pituch & Stevens, 2015) and Analysis of Moment Structure 25.0 (AMOS; Arbuckle, 2011) software were used. Following the principles of testing a quantitative measure in the context of the dual-frame sampling method, the process of validating the quantitative measure (EFA, CFA and MGCFA) was conducted on sample A (e.g., the training sample) composed of a large sample size. Then, sample B (e.g., testing sample) was used to compute the reliability of War Child Agency Assessment Scale - Palestinian version (WCAAS-Pal) scores along with both convergent and divergent validity. For convergent validity (i.e., convergent validity establishes construct validity using different measurement procedures to collect data about the same construct; Campbell & Fiske, 1959), WCAAS-Pal scores correlated with Children Hope Scale measures. For divergent validity (i.e., divergent validity establishes construct validity by supporting the idea that the construct of interest is different from other constructs that might be present in the study), WCAAS-Pal scores were correlated with Children Revised Impact of Event Scale-Arabic Version measures (e.g., psychological symptoms related to the exposure of potentially traumatic events).