ANALYSIS OF THE EFFECT OF LEADERSHIP, COMMUNICATION, AND MOTIVATION ON TURNOVER INTENTION WITH INTERVENING COMMITMENT ORGANIZATION VARIABLES AT PT MEDIANUSA PERMANA, BATAM CITY

The purpose of this study was to determine and analyze the effect of leadership, communication, and motivation on turnover intention with intervening commitment to organizational variables at PT Medianusa Permana, Batam City. The method used is a questionnaire and distributed to 100 respondents. Analysis of statistical data using SEM PLS (Structural Equation Modeling Partial Least Square) and using path analysis to examine the pattern of the relationship between the influence of the dependent variable on the independent, both direct and indirect effects with SMART PLS 3.0 software. The results in the study indicate that leadership has a positive and significant effect on organizational commitment with a p-value of 0.000 <0.05, communication has a positive and significant effect on organizational commitment with a p-value of 0.000 <0.05,turnover intentionwithp-value of 0.771,leadership has a positive and significant effect onturnover intentionwithp-value of 0.008, communicationleadership has a positive and significant effect onturnover intentionwithp-value of 0.004, motivationalhas a positive and significant effect onturnover intentionwithp-value of 0.018,organizational commitment mediates the effect of leadership andturnover intentionwith a p-value of 0.080 < 0.05. organizational commitment mediates the effect of communication andturnover intentionwith a p-value of 0.080 <0.05, and organizational commitment mediates the effect of motivation andturnover intentionwith a p-value of 0.080 < 0.05.

From the results of observations and interviews with several employees, the increase in the incidence of employee turnover intention at PT Medianusa Permana in 2017-2021 was caused by the leadership applying an inappropriate leadership style, such as not being close or limiting the distance to employees and low communication and employee motivation. The leadership style seen at PT Medianusa Permana is that leaders pay little attention to the needs of their employees at work, leaders pay little attention to employee complaints, leaders are arbitrary in assigning tasks, and leaders tend to be subjective in promoting employee positions, and lack of communication with employees, and are judged less fair in taking policies related to employees.
Seeing the description described above, it is necessary to conduct research that raises the topic of leadership, communication, motivation and organizational commitment influencing the occurrence of turnover intention. The author is interested in conducting research with the title "Analysis of the Influence of Leadership, Communication and Work Motivation on Employee Turnover Intention with Organizational Commitment as a Mediation Variable in PT. Medianusa Permana.

PROBLEM FORMULATION
The formulation of the problem is a problem questions that seek answers through data collection. The formulation of the problem in this study is:

RESEARCH METHODOLOGY
In this study, using respondent data, such as gender, age, education, and length of work in order to provide information about the characteristics of the respondents. A total of 100 questionnaires were distributed. The discussion in this chapter is the result of a field study to obtain questionnaire answer data that measures the six main research variables, namelyleadership, communication, motivation, organizational commitment, and Turnover Intention. Analysis of data with parametric and non-parametric statistics using SEM-PLS (Structural Equation Modeling-Partial Least Square) regarding research variables, validity and reliability as well as discussion of the results of hypothesis testing. This study uses path analysis (path analysis) to examine the relationship pattern of the influence of a variable or set of variables on other variables, both direct and indirect effects. Calculation of the path coefficient in this study was assisted by Smart PLS Ver 3.0. To find out the direct and indirect effects between variables, it can be seen from the calculation of the path coefficient to find out the significance.
The population in this study were the employees of the Regional Secretariat of Karimun Regency, totaling 138 people. PThis research uses the census method by giving questionnaires to the entire population. All members of the population used as a sample is called saturated sampling or another term is a census.

Outer Model Testing (Measurement Model)
This research model will be analyzed using the Partial Least Square (PLS) method and assisted with SmartPLS 3.0 software. PLS is an alternative method of Structural Equation Modeling (SEM). Outer model testing consists of convergent validity and discriminant validity.

a. Convergent Validity
Convergent validity is used to see the extent to which a measurement is positively correlated with alternative measurements of the same construct. To see an indicator of a construct variable is valid or not, then look at the value of the outer loading. The outer loading test with a value greater than 0.5 given in the SMARTPLS procedure is generally considered significant (Hair et. al, 2014). In the early stages of outer loading testing, all indicators with external load values were above 0.5, therefore, these indicators were valid and included in this study. The following results of the loading factor can be shown as in From the results of data processing with SmartPLS shown in Table 1, that the majority of the 64 indicators for each variable in this study have a loading factor value greater than 0.5 and are said to be valid. The next evaluation is by comparing the AVE root value with the correlation between the constructs. The recommended result is that the AVE root value must be higher than the correlation between the constructs (Yamin and Kurniawan, 2011). The model has better discriminant validity if the AVE square root for each construct is greater than the correlation between the two constructs in the model. A good AVE value is required to have a value greater than 0.50. In this study, the AVE value and the AVE square root for each construct can be shown in Table 2: Based on the AVE value in table 2, it shows that the AVE value is greater than 0.50, with the smallest value being 0.513 for the Communication variable (X2) and the largest being 0.597 for the Organizational Commitment variable (X1). This value meets the requirements in accordance with the specified minimum AVE value limit of 0.50. After knowing the AVE value for each construct, the next step is to look at the VIF value.
The collinearity test is to prove whether the correlation between latent/construct variables is strong or not. The value used to analyze it is by looking at the Variance Inflation Factor (VIF) value (Hair, et. al 2014;Garson, 2016). If the VIF value > 5.00, it means that there is a collinearity problem, and conversely there is no collinearity problem if the VIF value is < 5.00. From the data above it can be described as follows: 1) VIF for the correlation X1 with Y is 3,318 < 5.00 (no collinearity problem) 2) VIF for the correlation X2 with Y is 4,082 < 5.00 (no collinearity problem) 3) VIF for the correlation X3 with Y is 2,468 < 5.00 (no collinearity problem) 4) VIF for correlation Z with Y is 4.166 < 5.00 (no collinearity problem) Thus, from the data above, the structural model in this case does not contain collinearity problems.
Outer modelBesides being measured by assessing convergent validity and discriminant validity, it can also be measured by looking at the reliability of constructs or latent variables as measured by composite reliability values. The construct is declared reliable if the composite reliability has a value > 0.7, then the construct is declared reliable. The SmartPLS output results for composite reliability values can be shown in Table 4. From the SmartPLS output results in Table 4, the composite reliability value for all constructs is above the value of 0.70. With the resulting value, all constructs have good reliability in accordance with the minimum value limit that has been required.

4.2.Inner Model Testing (Structural Model)
The inner model (inner relation, structural model, and substantive theory) describes the relationship between latent variables based on substantive theory. The inner model can be evaluated by looking at the r-square (indicator reliability) for the dependent construct and the tstatistical value of the path coefficient test. The higher the r-square value means the better the prediction model of the proposed research model. The path coefficients value indicates the level of significance in hypothesis testing.
The structural model is evaluated using R-square for the dependent construct. The R² value can be used to assess the effect of certain endogenous variables and whether exogenous variables have a substantive effect (Ghozali, 2014). Based on table 5, the R Square value of the Organizational Commitment variable is 0.760, this means that 76.0% of variations or changes in organizational commitment are influenced by leadership, communication, and work motivation while the remaining 24.0% is explained by other reasons. The R Square variable Turn Over Intention is 0.733, this means that 73.3% of the variation or change in Turn Over Intention is influenced by organizational commitment while the remaining 26.70% is explained by other reasons. So it can be said that the R Square on the Consumer Satisfaction variable is moderate.
Hypothesis testing is carried out based on the results of testing the Inner Model (structural model) which includes the output r-square, parameter coefficients and t-statistics. To see whether a hypothesis can be accepted or rejected by considering the significance value between constructs, tstatistics, and p-values. Testing the research hypothesis was carried out with the help of SmartPLS (Partial Least Square) 3.0 software. These values can be seen from the bootstrapping results. The rules of thumb used in this study are the t-statistic >1.96 with a significance level of p-value 0.05 (5%) and the beta coefficient is positive. The value of testing the hypothesis of this study can be shown in Table 6, Table 7    Based on the table above, the analysis obtained; 1. The direct effect of variable X1 on variable Z has a path coefficient of 3,582 (positive), has a P-Values of 0.000 <0.05, so it can be stated that the influence between X1 on Z is significant. 2. The direct effect of variable X2 on variable Z has a path coefficient of 4,673 (positive), a P-Values of 0.000 <0.05, so it can be stated that the influence of X2 on Z is significant. 3. The direct effect of variable X3 on variable Z has a path coefficient of 2,455 (positive), a Pvalue of 0.014 <0.05, so it can be stated that the influence of X3 on Z is significant. 4. The direct effect of variable Z on variable Y has a path coefficient of 0.638 (positive), a Pvalue of 0.771 <0.05, so it can be stated that the influence of Z on Y is not significant. 5. The direct effect of variable X1 on variable Y has a path coefficient of 2.668 (positive), a P-Values of 0.008 <0.05, so it can be stated that the influence between X1 on Y is significant. 6. The direct effect of variable X2 on variable Y has a path coefficient of 2,707 (positive), a P-Values of 0.004 <0.05, so it can be stated that the influence of X2 on Y is significant. 7. The direct effect of variable X3 on variable Y has a path coefficient of 2,839 (positive), a P-Values of 0.018 <0.05, so it can be stated that the influence of X3 on Y is significant.
Testing the indirect effect hypothesis aims to prove the hypotheses of the effect of a variable on other variables indirectly (through an intermediary). If the value of the coefficient of indirect effect > the coefficient of direct effect, then the intervening variable mediates the relationship between one variable and another. Conversely, if the indirect effect coefficient < direct effect coefficient, then the intervening variable does not mediate the relationship between one variable and another variable.  (2), 8-11. https://medium.com/@arifwicaksanaa/pengertian-use-case-a7e576e1b6bf Friska Sihombing, E., Satriawan, B., Indrayani, I.,