*3.4. Procedure and Data Analysis*

The online questionnaire was sent to all FSC participants in the period from August– September 2022. Wholesalers, independent carriers, and retailers, as well as the largest agricultural holdings, transport and logistics centers, and retail chains in the territory of the western Balkans (Serbia, Croatia, B&H, Montenegro, North Macedonia), are equally represented in the sample. The sample consists of business entities that primarily deal with fast-moving consumer goods, except for fresh fruits and vegetables, fresh meat, fish, and other products that are marketed unpackaged or in bulk.

The questionnaire consisted of 22 questions that were structured based on similar questionnaires and research conducted in some earlier studies [14,16,18,28]. The questionnaire had three parts. After collecting general demographic information about the respondents (gender, age, and position), in the first part of the questionnaire, respondents were asked to evaluate the impact of each of the offered indicators on the functioning of the FSC. The indicators were evaluated based on three Likert-type items (0–5 scale). After that, in the second part, respondents ranked how implemented modern technologies (digitalization) minimized the negative impact of critical indicators. Digitalization was also operationalized through three items: (1) application of information technology: BT, IoT, DLT, etc.; (2) application of sensor and identification technology: WSN, TTI, Barcode, RFID, etc.; (3) application of location-based technology: RS, GPS, RTLS, etc. In the last part of the questionnaire, the direct impact of digitalization on the sustainability and functioning of the FSC was also assessed through three Likert-type items.

The total number of sent questionnaires was 600, which shows a return rate of filled questionnaires of 33.7% (242/640). The collected data were analyzed and used to test research hypotheses. The method of descriptive statistics was used to present the most significant characteristics of the sample, whereas the basic and supporting research hypotheses were tested using the statistical method of structural modeling (SEM) or path analysis.

IBM SPSS Amos 23 structural equation modeling software was used for data design and analysis. Path coefficients (R*ij*) were calculated programmatically based on the following pattern:

$$\text{Rij} = \text{Pi}\dot{\text{j}} + \sum (\text{R } ik \times \text{P } k\text{j}) \tag{1}$$

wherein:

R*ij*—the mutual connection between independent indicators (i) and dependent variables (j) measured by the correlation coefficient (r),

P*ij*—the component that shows the direct influence (effect) of independent indicators (i) on the dependent variable (j) measured by the path coefficient,

∑(R *ik* × P *kj*)—the sum of the components of the indirect influence of a given independent indicator (i) on a given dependent variable (j) through independent characters (k).

The residual effect is determined based on the formula <sup>√</sup> 1 −R2, where—R<sup>2</sup> = ∑(R*ij* ×P *ij*).

To evaluate the model, that is, whether there is enough information to calculate unknown parameters in SEM, the following coefficients were used: NFI—Bentler-Bonett Normed Fit Index, RFI—Relative Fit Index, IFI—Incremental Fit Index, CFI—Comparative fit index, TLI—Tucker-Lewis index, RMSEA—Root Mean Square Error of Approximation, and CMIN/DF—Chi-square value/degree of freedom.

Other used statistical indicators were: Standard Error (SE = SD/√ n, where SD is the standard deviation and n is the number of elements in the sample), Standard Deviation (SD = p 1/N ∑(xi − *µ* ) 2 , where N is the number of elements in the sample, xi is the ith member of the sample, and µ is the arithmetic mean), Coefficient Beta (β = (Sx/Sy)b, where S<sup>x</sup> is the standard deviation of variable x, S<sup>y</sup> is the standard deviation variables y, and b is the standard regression coefficient), T value (t = ( x − µ)/(SD − √ n), where x is the arithmetic mean of the sample, µ is the arithmetic mean of the population, SD is the standard deviation of the sample, and n is the sample size).
