A Standard Scientific Research Process for a Quantitative Thesis

What is the scientific research process in a thesis

The scientific research process is a sequence of tasks that connects a real-world problem to a data-based answer. For a quantitative thesis, it usually includes choosing a problem, defining the objectives and research questions, reviewing the literature, building the model and hypotheses, designing the scale, collecting data, cleaning the data, running the analysis, interpreting the results, and writing the thesis.

You should treat the process as a chain of concrete outputs. After each stage, you need something you can check: the topic and research questions, the conceptual model, the questionnaire, the .sav or .csv data file, the SPSS or SmartPLS output, and the written sections of each chapter. If one stage does not produce a clear output, the next stage usually suffers and you end up revising several parts at once.

This article focuses on quantitative research using questionnaires, Likert scales, SPSS, or SmartPLS. You can also read the overview of scientific research methods if you need to distinguish between a research method, research design, and analysis technique.

The process is not a separate chapter with a fixed page count. Its stages are usually distributed across the literature review, methodology, results, and conclusion chapters. In a thesis of about 50 to 80 pages, the methodology chapter usually needs enough detail for the reader to understand how you sampled, measured, and analyzed the data.

Standard structure of the scientific research process

The table below gives you a quick check of what each stage should contain. The page counts are only reference ranges because each department may have different requirements.

You can also read research methods to name your research design correctly. Do not write “quantitative research” as though it describes the entire method. You still need to state whether the design is cross-sectional or longitudinal, how you sampled, what you measured, and which analysis technique you used.

Stages of building a scientific research process

Stage 1: Finalize the problem, topic, and scope

Start with a problem that can be measured using data. Instead of writing “a study of online shopping behavior,” for example, you could narrow it to “factors affecting the online shopping intention of students in Ho Chi Minh City.” This title already indicates the participants, behavior, and context.

Before finalizing the topic, check three points: whether you can reach the respondents, whether suitable scales exist, and whether the data collection period fits your submission schedule. The Undergraduate Thesis Topics page can help you compare possible directions, but the final topic must match the data you can actually collect.

Stage 2: Write the objectives and research questions

The overall objective describes what you want to explain. Specific objectives should connect to actions you can carry out, such as identifying influencing factors, measuring the size of their effects, and proposing managerial implications.

Your research questions must match your objectives. If you ask which factors have an effect, the model needs independent variables. If you ask about the size of the effect, you need an appropriate regression model or PLS-SEM. If the question concerns mediation or moderation, the questionnaire and analysis plan must also be designed for that relationship.

Stage 3: Review the literature and identify the gap

Create a table with the author, year, context, sample, variables, method, and main findings. This helps you see which parts previous studies have answered and how your topic differs in its participants, context, variables, or model.

When searching for sources, do not collect only papers with similar titles. Read the definitions of the concepts, the sources of the scales, and the way the model was tested. If you need foundational material in Vietnamese, scientific research methods textbook PDF is a starting point, but each specific scale still needs to be traced to an appropriate original source.

Stage 4: Build the model and hypotheses

Every arrow in the model needs a reason based on theory or previous findings. State the independent variables, dependent variables, and any mediator or moderator. Then write each hypothesis as a complete sentence, for example: “H1: Perceived usefulness has a positive effect on students’ intention to use a learning application.”

Before moving to questionnaire design, check the expected number of items, the direction of each effect, and how each construct will be measured. A model with too many variables makes the questionnaire long, increases the risk that respondents will abandon it, and makes the results harder to explain.

Stage 5: Select the scales and design the questionnaire

Find the original scale for each construct and then adjust the wording to fit the Vietnamese context. Each item needs its own code, such as PU1, PU2, or SAT1, SAT2. If you have reverse-coded items, note them during the design stage so that you can recode them before running the analysis.

Keep a 5-point or 7-point Likert scale consistent throughout the questionnaire. The opening section should contain screening questions, the main section should contain the items, and the final section should collect descriptive information about the sample. You can refer to quantitative research methods to check whether your questions, scales, and model fit together.

Stage 6: Define the sample and collect the data

State the target population, respondent criteria, sampling method, collection period, and collection channel. Explain the sample size using the number of items, number of independent variables, or population size instead of writing a number simply because “many sources use it.” Hair et al. suggest using 5 to 10 observations per item as a reference (Hair et al., 2010). For regression, a commonly used rule is n of at least 50 + 8m, where m is the number of independent variables (Tabachnick and Fidell, 2013).

Before distributing the official questionnaire, check how respondents understand the wording, how long completion takes, and whether required responses work correctly. During collection, keep the raw data file, cleaned file, and analysis file separate. If you use DoThesis, fillform.info is our form product for collecting questionnaire responses in Vietnamese. Google Forms is a familiar alternative.

Stage 7: Clean and analyze the data

Check for duplicate records, blank responses, unrealistically short completion times, respondents selecting the same option for nearly every item, and values outside the specified codes. Record every decision to remove a record in a log file. Do not delete records directly without leaving an audit trail.

The analysis sequence usually includes descriptive statistics, Cronbach's Alpha, EFA if the model requires it, and then regression or PLS-SEM. For EFA, KMO of 0.5 or above and a statistically significant Bartlett’s test are commonly referenced conditions (Kaiser, 1974). A factor loading of 0.5 or above and total variance explained of 50% or above are often used as reference points (Hair et al., 2010). Do not automatically remove an item because one number looks weak. Review the scale content and model structure as well.

Stage 8: Write the conclusions, implications, and final thesis

After obtaining the results, return to each research question and hypothesis. Whether a hypothesis is supported must be based on the coefficient, direction of the effect, and significance level in the output, not on what you expected at the beginning.

Managerial implications should follow from the factors that have an effect and from the research context. State limitations specifically, including the sampling method, geographical scope, cross-sectional design, number of variables, or limits on generalizability. Also check that the variable names match across the model, questionnaire, data file, and tables in Chapter 4.

Usable sample sentences

The sentences below are templates. Replace the bracketed sections with real information. Do not copy a sentence if it does not match your data.

How to write this in your thesis

You can write: “The research process consisted of identifying the problem, developing the model and hypotheses, designing the scale, collecting and cleaning the data, analyzing the data, and then discussing the results in relation to the research objectives.” Replace “the stages” with the exact sequence you followed, especially if you pilot-tested the questionnaire or adjusted items.

When reporting the sample size, you can write: “The study obtained [number] valid responses from [participants] after excluding [number] responses that did not meet [exclusion criterion].” This sentence forces you to reconcile the numbers in the questionnaire, data file, and results chapter.

A shortened example of a scientific research process

Suppose you are studying the factors affecting students’ intention to continue using an online learning application. The problem is narrowed to four independent variables: perceived usefulness, perceived ease of use, content quality, and social influence. Intention to continue using the application is the dependent variable.

You begin by finding the theoretical sources and scales, draw a model with four arrows pointing to intention to continue using the application, and write four corresponding hypotheses. The questionnaire uses a 5-point Likert scale, includes 20 items, and contains a demographic section. Under the rule of 5 to 10 observations per item (Hair et al., 2010), you can plan to collect approximately 100 to 200 responses, then compare that plan with the requirements of the model and your supervisor.

After collecting the data, code the variables, check for invalid records, run Cronbach's Alpha, and run EFA if appropriate. Next, run regression or PLS-SEM, review the effect coefficients and significance levels, and return to the four hypotheses. If you use PLS-SEM, the reporting process should separate the assessment of the measurement model from the structural model, consistent with the guidance in (Hair et al., 2022).

A shortened paragraph could read: “This study uses a quantitative method with data collected from students who have used an online learning application. The proposed model tests the effects of perceived usefulness, perceived ease of use, content quality, and social influence on intention to continue using the application. The data were processed through scale reliability checks, measurement validity assessment, and tests of the relationships in the model.” Keep only the sentences that match your actual software and procedure.

Common problems that make the committee ask follow-up questions

The first problem is that the topic, objectives, model, and questionnaire do not tell the same story. The title refers to purchase intention, the objectives refer to satisfaction, and the questionnaire measures loyalty. When questioned, you will not be able to explain which variable is the main outcome.

The second problem is citing the scale source without explaining how you adapted it. The committee may ask how the items were translated, whether any items were reverse-coded, and why some items were removed. Keep a comparison table showing the original scale, the Vietnamese version, and the version used in the survey.

The third problem is inconsistent response counts. The number of collected responses, excluded responses, and responses entered into SPSS must have a clear relationship. The final data file must also match the number of observations in the Descriptive Statistics, Reliability Statistics, and Model Summary tables.

The fourth problem is changing the results to fit the hypotheses. Deleting items, removing cases, or rerunning the analysis repeatedly just to obtain an acceptable p-value creates greater risk during the defense. If the EFA does not look clean on the first run, record why an item was removed, rerun the analysis using a defensible criterion, and report the process consistently.

The final problem is mixing criteria from different methods. CFI, TLI, and RMSEA are commonly used with CB-SEM, while SmartPLS requires measures such as outer loading, CR, AVE, HTMT, VIF, R², and Q² according to the analysis design. Choosing one software package and then applying another package’s criteria creates difficult questions during the defense.

Frequently asked questions

How many stages are in the scientific research process?

There is no single number of stages for every topic. For a quantitative thesis, you can organize the process into eight stages: defining the problem, setting the objectives and questions, reviewing the literature, building the model, designing the scale, collecting the data, analyzing the data, and then writing the conclusions and recommendations.

The important point is that every stage has an output and connects to the next stage. If your department requires a different number of stages, you can combine or separate them while keeping the necessary content.

How does the quantitative scientific research process differ from the qualitative process?

Quantitative research uses numerical data, items, questionnaires, and analyses such as Cronbach's Alpha, EFA, regression, or PLS-SEM. Qualitative research has different objectives and data collection methods, so you should not transfer its process into a thesis that uses SPSS.

If your topic uses only a questionnaire and a data file, state the quantitative design, sampling method, scale, and analysis criteria clearly. You can read the page on qualitative research methods to recognize the difference in scope, but you do not need to add techniques outside your design to the thesis.

Does the scientific research process require a pilot survey?

A pilot survey is useful when the wording is difficult to understand, the questionnaire is too long, or the coding is inconsistent. State that it was a tool-checking stage, separate the pilot data from the official data where possible, and do not combine the two types of data arbitrarily.

If your supervisor does not require a pilot survey, you should still check the questionnaire internally before distributing it. Report what you checked and what decision you made about each adjustment.

What sample size is sufficient for the scientific research process?

The required sample size depends on the number of items, number of independent variables, analysis model, population, and access to respondents. The reference point of 5 to 10 observations per item can be used for planning (Hair et al., 2010), while regression has the rule of n at least 50 + 8m (Tabachnick and Fidell, 2013).

State how you calculated the sample size and report the actual number of valid responses in the methodology chapter. A sample that exceeds the reference point can still be problematic if the data are unsuitable, responses are duplicated, or the sample does not match the target population.

If one stage is wrong, do you have to redo everything?

Not always. If the problem is a variable format or coding error, you can correct the file and rerun the relevant analysis. If the questionnaire measures the wrong construct or the sample does not match the target population, changing the software settings will not solve the problem.

Make an error list organized by impact, ask your supervisor about major decisions, and save each file version. This makes it easier to explain why the final results differ from an earlier run.

In which chapter should you write the scientific research process?

The stages are distributed across several chapters. Chapter 1 presents the problem, objectives, and questions. Chapter 2 presents the theoretical basis, previous studies, model, and hypotheses. Chapter 3 presents the design, sample, scales, and analysis method. Chapter 4 presents the results, and Chapter 5 presents the conclusions and implications.

You can add a process diagram at the beginning of the methodology chapter if your department’s template allows it. The diagram is useful only when its stages match what you actually did.

Open your outline file and data file now, create eight lines corresponding to the eight stages, and enter one concrete output and one thesis location for each line. If you need to run the analysis on your own .sav or .csv file, M4 analysis by DoThesis supports SPSS and SmartPLS workflows using real data.