*4.1. Pilot Test*

As our questionnaire items were adopted from previously published studies, we carried out a pilot test to guarantee the validity and reliability before a broad distribution of the questionnaires. The pilot test was carried out on a sample of 39 responses, with a minimum suitable threshold of thirty entries for the pilot run as stated by Johanson and Brooks (2010). To test the validity of items, the factor analysis (multivariate technique) was used. Data are subjected to two requirements prior to factor analysis: Kaiser Meyer-Olkin (KMO) test and Bartlett's test of sphericity. The KMO value for our sample is 0.899, which is substantially higher than a threshold value of 0.60 Kaiser (1974).

Bartlett's sphericity test determines if the correlation matrix is an identity matrix, indicating that the factor model is inapplicable Malhotra et al. (2006). If the Bartlett value is significant (*p* = 0.05), then it is possible to employ the principle component analysis (PCA), a process used to compress a larger collection of variables into smaller ones to optimize the interpretation and minimize the information loss Jolliffe and Cadima (2016). For our case, the null hypothesis was rejected (approx. chi-square: 1 221.46; DF = 210; *p* = 0.000) indicating that the variables in the population correlation matrix were uncorrelated. Table 1 contains the factor loadings for the pilot test (items with loadings at 0.40 or less were deleted from the final survey questionnaire).


**Table 1.** Measurement model evaluation for validity and reliability.

#### *4.2. Data Analysis*

Given that the purpose of this study is to predict employee performance based on the leaders' expectations and leader-member exchange, the partial least square structural equation modeling (PLS-SEM) technique was used. PLS-SEM is a common approach for management and social science research which has been widely applied in prior studies Aydin (2020); Khan et al. (2020); Khan (2022) when the major purpose of the study is to assess a core model and show a target construct Hair et al. (2019). Additionally, it prioritizes the optimization of the endogenous construct prophesy above the model fit.

The results were computed with the help of SmartPLS Software (3rd version). PLS-SEM is a two-step process. In the first stage, the measurement model's validity and reliability are evaluated. Table 1 displays the composite reliability (CR), extracted average variance (AVE) and factor loadings (FL). The observed values were above the 0.70 cut-off threshold for factor loadings and composite reliability, and above the 0.50 cut-off point for average variance Hair et al. (2017), thus confirming the measurement model's internal consistency and convergent validity.

We analyze the measurement model's discriminant validity (DV) using the Fornell-Larcker criteria and the heterotrait-monotrait (HTMT) correlation ratio, as indicated by Hair et al. (2017). The degree to which one construct differs from another is referred to as discriminant validity. The observed HTMT ratio for all variables was below 0.85; our model also meets the Fornell-Larcker criterion, demonstrating discriminant validity (see Table 2).

The structural model is examined in the second stage of PLS-SEM evaluation. The bootstrapping tool integrated in SmartPLS Software was used to assess the relevance of all path coefficients. The results displayed in the Table 3 demonstrate that all five hypothesized relationship were supported. The graphical representation of our outcome is depicted in Figure 2.


**Table 2.** Heterotrait-monotrait (HTMT) ratio and Fornell-Larcker criterion.

**Table 3.** Findings.


**Figure 2.** Summary of research framework.
