Research Design and Participants
The method of the current research was correlational of structural equation modeling type. The statistical population of the present study included all the college students of Kermanshah, Iran in the academic year 2020–2021, of whom 338 (256 women and 82 men) were selected through convenience sampling. It is worth explaining that Stevens (1996, as cited in 38) considered 15 cases for each predictor variable in multiple regression analysis with the standard least squares model as a good rule of thumb. Based on this issue, it can be stated that as SEM is completely related to multivariate regression in some aspects, the number of 15 items for each measured variable in SEM is not unreasonable. Loehlin (1992, as cited in 39) stated that for models with two or four factors, the researcher should plan on collecting at least 100 cases or more, for example, 200 cases. Therefore, in order to determine the sample size, the calculation based on the number of components of the research variables was used. Taking into account the number of samples of 15 for each component and considering that the current research included 13 components, the minimum number of samples needed to conduct the study was 195. However, since there is always the possibility of participants dropping out and having distorted questionnaires, and Loehlin’s recommendation to have a sample of more than 200, we tried to consider more than 300 people as the study sample, and finally 338 psubjects completed our questionnaires. The inclusion criteria were (1) studying in one of the universities of Kermanshah at any stage and (2) the desire to participate in the study, and the exclusion criterion was the experience of bereavement of a loved one in the last three months. The average age (18 to 53 years) of the participants was generally 25.09 (SE = 5.02). Among participants, 282 were single (83.4%) and 56 married (16.6%). Out of all participants, 10 (3%) reported their education as associate of art, 215 (63.6%) as bachelor, 104 (30.8%) as master and 9 (2.7%) as PhD student.
Measures
Social and Emotional Loneliness Scale for Adults (SELSA-S). This scale was designed and prepared by Ditommaso et al. (40). This scale includes 14 items and its three subscales include romantic loneliness, family loneliness and social loneliness. For each item, there is a 5-point Likert scale (1 = completely disagree to 5 = completely agree). The minimum possible score will be 14 and the maximum will be 70. A lower score means less loneliness in adults. Ditommaso et al. (40) reported its Cronbach’s alpha coefficient between 0.87 and 0.90, which indicates the appropriate internal consistency of the scale. They have also reported that there is a significant correlation between SELSA-S with the total score of Russell’s (1978, as cited in 41) revised loneliness scale. In Iran, confirmatory factor analysis of the questionnaire, done by Jowkar and Salimi (42), showed that the three-factor structure has a good fit with the data. Cronbach’s alpha coefficients for the subscales of romantic, social, and family loneliness were equal to 0.92, 0.84, and 0.78, respectively (42). In the present study, the reliability of this questionnaire based on Cronbach’s alpha for subscales of romantic, family and social lonliness, were 0.91, 0.89 and 0.81, respectively
Adult Attachment Questionnaire (AAQ). This questionnaire, created by Hazan and Shaver (20), distinguishs three attachment styles of secure, avoidant and ambivalent. This questionnaire has 21 items scored on a 5-point Likert scale (1 = none to 5 = very high). A higher score in each of the subscales means that the characteristics of that type of attachment are high in the person and vice versa. Hazen and Shaver found the reliability of this questionnaire to be 0.81 using retest method with a time interval of two weeks, and 0.78 with Cronbach’s alpha method. In Iran, Cronbach’s alpha coefficient of each of the security, avoidant and ambivalent subscales for a student sample were 0.74, 0.72, 0.72 respectively (Besharat, 2003, as cited in 43). In the study of Parvizi and Sadeghi (44) this questionnaire was subjected to confirmatory factor analysis, and the results revealed the good fit of the model with the data. In the present study, the reliability of this questionnaire based on Cronbach’s alpha for secure, ambivalent and avoidant subscales were obtained to be 0.81, 0.86, and 0.89, respectively.
Young Schema Questionnaire (YSQ-SF). This questionnaire, created by Young (45). This questionnaire has 75 items to identify 15 EMS in the form of five domains including emotional deprivation, abandonment/instability, mistrust/abuse, social isolation/alienation, defectiveness/shame, failure, dependence/incompetence, vulnerability to harm and illness, enmeshment/undeveloped self, subjugation, self-sacrifice, emotional inhibition, unrelenting standards/hypercriticalness, entitlement/grandiosity and insufficient self-control/self-discipline. Questions are scored on a 6-point Likert scale. A higher average in each schema means that schema is active in the person. Smith et al. (1995, as cited in 46) indicated that the Cronbach’s alpha coefficient in the non-clinical population for the subscales of this questionnaire was between 0.5 and 0.82. Using confirmatory factor analysis, Ghiasi et al. (47) showed that the factor structure of the questionnaire has a good fit with the data. They also found that Cronbach’s alpha coefficients for the disconnection-rejection and other-directedness domains were 0.90 and 0.76, respectively. In this research, only two domains of disconnection-rejection (including schemas of emotional deprivation, abandonment/instability, mistrust/abuse, social isolation/alienation and defectiveness/shame) and other-directedness (including schemas of subjugation and self-sacrifice) were used. In the present study, the reliability of this questionnaire was obtained based on Cronbach’s alpha for the subscales of disconnection-rejection and other-directedness to be 0.86 and 0.81 respectively.
Procedure
Having obtained the necessary permits to conduct this research from the competent authorities of Razi University of Kermanshah, we carried out the steps of data collection. In these stages, the necessary questionnaires were designed in the form of online questionnaires. Due to the simultaneous implementation of this study with the severe peak of the Covid-19 virus in Iran, using the convenience sampling method, the link of the questionnaire was provided to the college students studying in the universities of Kermanshah through e-mail or social networks. Before the participants complete the questionnaires, the objectives of the research were clarified for them. In this regard, explanations were provided regarding the disclosure of identity information, confidentiality and privacy of the participants, and their informed consent to participate in the research was obtained. First, the participants answered preliminary questions that included the inclusion criteria, and those who met the inclusion criteria were allowed access to the main questionnaires. In addition, in the online implementation of this research, the questionnaires were designed in such a way that the participants could send the questionnaire if they answered all the questions. Therefore, there were no missing data in the study. Finally, by examining 338 questionnaires completed by the statistical sample, after discarding 19 participants (due to the presence of outliers), the information obtained from 319 people was analyzed as the final sample of the research. The data obtained from the questionnaires were analyzed using the Pearson’s correlation coefficient method and structural equation modeling (SEM) in Lisrel 8.8 and SPSS-22 software.
Findings
Before performing the analysis, the assumptions of SEM including normality of distribution, independence of errors and multiple collinearity were checked. To assume the normality of the research variables, the skewness and kurtosis of the distribution of scores were used, and the results showed that the distribution of the scores of all variables is normal (the range of distribution was between + 1 and − 1). Durbin-Watson test was used to check the independence of errors, the results of which showed no correlation between errors (DW = 1.98, the range between 1.5 and 2.5 is acceptable). Variance Inflation Factor (VIF) and Tolerance were used to check the multiple collinearity between the predictor variables, and the results showed that there is non-collinearity between the variables (the range of VIF was less than 5 and the tolerance was higher than 0.1). The results related to these assumptions are presented in Table 1 along with the average and standard deviation of the variables.
Table 1
The mean, standard deviation, and information regarding assumptions of SEM
Variable | minimum | maximum | M | SD | skewness | kurtosis |
AS | | | | | | |
SE | 8 | 25 | 15.95 | 3.09 | .075 | .052 |
AM | 5 | 23 | 12.10 | 3.74 | .383 | .058 |
AV | 6 | 28 | 15.35 | 3.87 | .328 | .254 |
DRS | 20 | 108 | 49.5 | 16.13 | .839 | .613 |
ED | 4 | 24 | 10.97 | 5.17 | .557 | .564 |
AB | 4 | 24 | 11.7 | 4.81 | .649 | .240 |
MA | 4 | 24 | 10.6 | 4.48 | .657 | .036 |
SI | 4 | 24 | 9.53 | 4.31 | .808 | .025 |
DS | 4 | 20 | 6.95 | 3.47 | 1.306 | .980 |
ODS | 6 | 33 | 14.62 | 5.25 | .683 | .195 |
SU | 3 | 17 | 5.31 | 2.92 | 1.518 | 1.758 |
SS | 3 | 18 | 9.31 | 3.52 | .301 | − .554 |
LO | 11 | 54 | 26.06 | 8.74 | .406 | − .073 |
RO | 3 | 15 | 8.89 | 1.4 | .103 | -1.202 |
FA | 4 | 20 | 8.06 | 3.7 | .944 | .38 |
SO | 4 | 20 | 9.14 | 3.70 | .716 | .032 |
Note. AB = Abandonment/Instability, AM = Ambivalent, AS = Attachment Styles, AV = Avoidant, DRS = Disconnection-rejection Schema, DS = Defectiveness/Shame, ED = Emotional Deprivation, FA = Family, LO = Loneliness, MA = Mistrust/Abuse, ODS = Other- Directedness Schema, RO = Romantic, SI = Social Isolation/Alienation, SO = Social, SS = Self-Sacrifice, SU = Subjugation |
----Table 1----
Another assumption is the presence of a linear relationship between independent and dependent variables, which was investigated with Pearson’s correlation coefficient, the results of which are reported in Table 2.
Table 2
The correlation coefficient between the variables
variable | LO | RO | FA | SO |
SE | − .196** | − .082 | .031 | .268** |
AM | .118* | .069 | .172** | .03 |
AV | .375** | .287** | .283** | .383** |
DRS | .505** | .346** | .413** | .394** |
ED | .571** | .436** | .447** | .416** |
AB | .086** | .027 | .091 | .082 |
MA | .268** | .203** | .213** | .194** |
SI | .456** | .288** | .375** | .381** |
DS | .466** | .303** | .388** | .374** |
ODS | .148** | .071 | .177** | .094** |
SU | .226** | .061 | .282** | .183** |
SS | .033 | .055 | .029 | .012** |
Note. *P < .05, **P < .01 Note. AB = Abandonment/Instability, AM = Ambivalent, AS = Attachment Styles, AV = Avoidant, DRS = Disconnection-rejection Schema, DS = Defectiveness/Shame, ED = Emotional Deprivation, FA = Family, LO = Loneliness, MA = Mistrust/Abuse, ODS = Other-Directedness Schema, RO = Romantic, SI = Social Isolation/Alienation, SO = Social, SS = Self- Sacrifice, SU = Subjugation |
----Table 2----
Table 3 shows the fit indices of the proposed and modified research model. According to the contents of Table 3, to determine the adequacy of the proposed model with the data, a combination of goodness-of-fit indices such as chi-square value (χ2), normalized chi-square measure (the ratio of chi-square to degrees of freedom), goodness-of-fit indices (GFI), normalized fit (NFI), comparative fit (CFI) and root mean square error of approximation (RMSEA) were used.
Table 3
Fit indices for the developed model
Model fit indices | X2 | df | X2/ df | GFI | IFI | CFI | RMSEA |
Obtained values | 917.96 | 334 | 2.74 | 0.93 | 0.92 | 0.91 | 0.075 |
-
----Figure2----
-
----Table 3----
According to Table 3, the amount of fit indices such as incremental fit index (IFI = 0.92), comparative fit index (CFI = 0.91), normalized fit index (NFI = 0.91) as well as root mean square of the approximate residuals (RMSEA = 0.075) indicates a very good fit of the model with the data. Other fit indices such as goodness of fit index (GFI = 0.93) and adjusted goodness of fit index (AGFI = 0.92) are also acceptable. Table 4 shows the parameters related to the direct effects of variables on each other in the proposed model.
Table 4
Coefficients of the model of the relationship between AS (AM and AV) and EMS (DRS and ODS) through EMS and LO
Direct path | Regression coefficient | statistic- t |
AM → DRS | 0.81 | 6.11** |
AM → ODS | 0.93 | 6.77** |
AV → DRS | 0.47 | 4.33** |
AV → ODS | 0.27 | 3.35** |
DRS → LO | 2.1 | 4.56** |
ODS → LO | -1.5 | -3.73** |
**P < 0.01 Note. AM = Ambivalent, AS = Attachment Style, AV = Avoidant, DRS = Disconnection-rejection Schema, EMS = Early Maladaptive schema LO = Loneliness, ODS = Other- Directedness Schema |
----Table 4----
As the results in Table 4 show, the direct paths are significant and the path of secure attachment to the schemas of disconnection-rejection, other-directedness and loneliness is removed. Also, the Sobel test was used to investigate the mediating role of disconnection-rejection and other-directedness schemas in the relationship between attachment styles and loneliness, the results of which are reported in Table 5.
Table 5
Results of investigating the mediating role of DRS and ODS in the relationship between AS (AM and AV) and LO
Variables | P | Sobel’s test (z) |
AM → DRS → LO | P < 0.001 | 7.44 |
AM → ODS → LO | 0.01 | 2.6 |
AV → DRS → LO | P < 0.001 | 5.46 |
AV→ ODS → LO | 0.12 | 1.51 |
Note. AM = Ambivalent, AS = Attachment Style, AV = Avoidant, DRS = Disconnection-rejection Schema, LO = Loneliness, ODS = Other-Directedness Schema |
----Table 5----
Referring to Table 3, the proposed model of the research has good fit indicators. The root mean square index of the approximate residuals (RMSEA = 0.075) indicates a good fit of the model with the data. Therefore, the proposed model has a good fit in the target sample. In the meantime, secure attachment style could not play a role in the proposed model and was removed from the equations for better modeling, and two attachment styles, avoidant and ambivalent, remained in the model. According to the information in Table 5, the significance of the Sobel test indicates the significance of the mediation role of the disconnection-rejection schema and the other-directedness in the relationship between ambivalent attachment and loneliness. Avoidant attachment, however, only has a significant relationship with the loneliness through the schema of disconnection-rejection, and this is not significant through the mediating role of other-directedness schema.