10Min Research - 37. What to do after the Data Collection: How to Start the Data Analysis?

Once questionnaire data is collected, the next decisive step is setting up a clean analysis workflow—starting with coding, moving through descriptive reporting, and then running structural equation modeling (SEM) in either SmartPLS or IBM SPSS AMOS. The core idea is that analysis isn’t just “run the software”; it’s a sequence of checks and reporting requirements that make the measurement model credible before any hypothesis testing happens.

The process begins with coding every questionnaire item into a consistent variable naming scheme. For example, ethical responsibility items are coded as ER1–ER7, research and development responsibilities as RDR1–RDR6, and the same pattern is applied to each construct. Demographic variables are handled as separate coded variables as well. This coding step matters because it determines how constructs and indicators will be recognized later in SEM.

Next comes data preparation and quality screening. The data can be defined and entered in Excel or IBM SPSS: in SPSS, variables must be created in Variable View, while in Excel, column names are placed in the top row. After the dataset is in place, the workflow shifts to “screening” the data—checking minimum and maximum values, and identifying missing data—so that later reliability and validity tests aren’t distorted by errors or incomplete responses.

After cleaning, descriptive statistics are reported. This includes the demographic profile of respondents and descriptive summaries for indicators. The guidance here is practical for thesis writing: descriptive statistics for indicators may be omitted in the main paper if space is tight, but they can still be presented in a table.

With the dataset ready, the SEM section is structured around two layers: the measurement model and the structural model. In the measurement model, the analysis first reports factor loadings, then establishes construct reliability and construct validity. Reliability is assessed using Cronbach’s Alpha and composite reliability. Validity is assessed through convergent validity and discriminant validity. Only after these quality criteria are established does the report move to the structural model.

For the structural model, the reporting starts with explanatory power using R². Hypothesis testing follows: direct relationships are reported plainly, while mediation and moderation require additional model specifications. If IBM SPSS AMOS is used, the write-up must also include model fit for both the measurement and structural models. If SmartPLS is used, the reporting structure follows the standard SmartPLS approach.

Finally, the results chapter ends with a chapter summary. The emphasis throughout is on organizing the results section so that readers can see the logic: code and clean the data, describe it, validate the measurement model, then test the relationships with SEM using the appropriate software.