What Is Scientific Research? Process, Purpose, and How to Conduct It
What Is Scientific Research
Scientific research is a purposeful process for answering a research question with data, reasoning, and methods that can be checked. When you write a thesis, you start with a specific problem, identify the research subject, develop questions or hypotheses, collect data, and analyse it before reaching a supported conclusion.
The core requirement is a systematic process. You do not simply state a personal opinion about a phenomenon. You need to identify what must be explained, how it will be measured, where the data come from, and which rules you will use to draw conclusions. In quantitative research, data are usually entered from a questionnaire into a .sav or .csv file and then processed with SPSS, SmartPLS, or another suitable statistics program.
A scientific research project usually contains linked components: the research problem, objectives, questions, model, hypotheses, scales, research sample, data collection method, and analysis method. If one link does not fit, the final thesis becomes difficult to defend. For example, a hypothesis may test an effect while the questionnaire does not actually measure the variable involved.
You can read the short explanation of what scientific research is, then compare it with the research subject in your own project. Students often mix up these two concepts when they are starting.
The Role of Scientific Research in Quantitative Research
In quantitative research, scientific research helps you turn a broad statement into measurable variables. For example, “service quality affects satisfaction” is only a direction for the study. To test it, you need to identify the dimensions of service quality, determine which items measure satisfaction, and specify which data will show whether the relationship is supported.
Scientific research also helps you decide which analysis to run. If your objective is to assess scale reliability, you can run Cronbach's Alpha. If you need to explore the structure of groups of variables, EFA is more appropriate. If your model contains latent variables and effect paths, you can consider SmartPLS or AMOS depending on the research design.
A quantitative process usually follows this order:
- Choose the problem and limit the scope of the research.
- Read the literature to identify concepts, models, and gaps.
- Develop research questions or hypotheses.
- Choose the scales and design the questionnaire.
- Collect, code, and clean the data.
- Analyse the data in the order specified in the method section.
- Interpret and discuss the results, then state the implications.
You can also read what quantitative research is if you are still distinguishing numerical data, independent variables, dependent variables, and control variables. If your project uses a questionnaire, identifying the research field early will help you limit the sample and context instead of surveying an excessively broad group.
The practical value of a result does not come from having many numbers in the output. The result must answer the original question. A coefficient with a small p-value still needs to be interpreted in the context of the model, scale, sample size, and direction of the effect. Record the reason for choosing each analysis before opening SPSS, rather than running every table and trying to explain them afterward.
When Is Scientific Research Acceptable
There is no single number that determines whether scientific research is acceptable. Each stage has its own criteria. In a quantitative thesis, you need to check data suitability, scale reliability and validity, model quality, and then the hypotheses. The thresholds below are commonly used reference points, not reasons to delete variables mechanically.
| Item to check | Common threshold | Canonical source |
|---|---|---|
| Cronbach's Alpha | 0.70 or above | (Nunnally, 1978) |
| Alpha for a new or exploratory scale | May be 0.60 or above | (Hair et al., 2010) |
| Corrected Item-Total Correlation | 0.30 or above | (Nunnally and Bernstein, 1994) |
| KMO | 0.50 or above | (Kaiser, 1974) |
| Bartlett's Test | p-value below 0.05 | (Kaiser, 1974) |
| Eigenvalue for factor extraction | Above 1 | (Kaiser, 1960) |
| Total Variance Explained | 50% or above | (Hair et al., 2010) |
| Factor loading in EFA | 0.50 or above | (Hair et al., 2010) |
| Observations per item | About 5 to 10 | (Hair et al., 2010) |
Read these thresholds together with the research context. For example, an increase in Alpha after deleting an item does not by itself prove that the scale is better. The item may represent a theoretically meaningful aspect of the construct. Review the item wording, Corrected Item-Total Correlation, correlation matrix, and purpose of the scale as well.
If you run regression, sample size also needs to be considered in relation to the number of independent variables. A commonly cited rule sets the minimum sample size at 50 plus 8 times the number of independent variables, according to (Tabachnick and Fidell, 2013). This is a reference point. Your actual sample size must also reflect the sampling method, valid response rate, and model.
How to Read Scientific Research in Output
In an SPSS file, you will not usually find a table called “Scientific Research.” You read each table to assess one part of the process. For a scale, open Analyze > Scale > Reliability Analysis. The Reliability Statistics table reports Cronbach's Alpha and the number of items included. The Item-Total Statistics table reports Corrected Item-Total Correlation and Alpha if Item Deleted.
The table below is illustrative output, not the result of a real study. The column names are kept as SPSS usually displays them so you can compare them with the file open on your screen.
| Output table | Column or row to read | Illustrative data | Interpretation |
|---|---|---|---|
| Reliability Statistics | Cronbach's Alpha | 0.824 | Internal consistency reaches the reference point |
| Reliability Statistics | N of Items | 4 | Four items were included in this run |
| Item-Total Statistics | Corrected Item-Total Correlation, DV1 | 0.612 | The item correlates with the total scale |
| Item-Total Statistics | Corrected Item-Total Correlation, DV2 | 0.241 | Review the wording and possibility of deleting the item |
| Item-Total Statistics | Cronbach's Alpha if Item Deleted, DV2 | 0.851 | Alpha increases if DV2 is deleted, but this alone is not enough evidence for deletion |
If you run EFA, the KMO and Bartlett's Test table reports KMO and Bartlett's test. The Total Variance Explained table helps you review the number of factors and the percentage of variance explained. The Rotated Component Matrix shows which factor each item loads on most strongly. Check for low loadings, cross-loadings, and items loading onto a group that does not fit the theoretical model.
If you run regression, the Model Summary table reports R Square, the ANOVA table reports the overall model fit test, and the Coefficients table usually contains B, Beta, t, and Sig. B is the unstandardized coefficient. Beta is more convenient when you compare the relative effects of variables. Sig. is commonly used to assess the statistical evidence for a coefficient.
Do not draw a conclusion from one Sig. column alone. Read the direction of the coefficient, review the confidence interval if available, check VIF, inspect the residuals, and compare the result with the hypothesis written before the analysis. If your project uses SmartPLS, the corresponding tables may include Outer Loadings, Construct Reliability and Validity, Discriminant Validity, and Path Coefficients. CFI, TLI, and RMSEA belong to CB-SEM assessment and should not be inserted into a guide for evaluating a PLS-SEM model.
What to Do When Scientific Research Does Not Meet the Criteria
When an indicator does not meet the criterion, work from the data toward the model. Do not delete an item immediately after the first run. EFA rarely produces a perfectly clean structure on the first attempt, but every deletion needs a reason that you record.
Check the Coding and Original Data
Open Variable View to check the variable type, value labels, missing values, and direction of reverse-coded items. A reverse-coded item entered in the wrong direction can sharply reduce its correlation with the other items in the scale. Also check blank rows, values outside the Likert scale, and records with the same answer throughout the questionnaire.
Check Each Item
Read Item-Total Statistics, the correlation matrix, and the wording of each item. If Corrected Item-Total Correlation is low, check whether the item measures the same concept, whether its wording may be misunderstood, and whether it was coded incorrectly. Delete an item only when statistical evidence is combined with a theoretical reason.
Rerun the Analysis and Save Each Version
Save the syntax file or record the order of item deletions, Alpha, KMO, total variance explained, and factor structure after each round. The committee may ask why an item was deleted. A short tracking table lets you answer with the process instead of trying to reconstruct it from several output files.
Consider Collecting More Data
If the problem comes from a sample that is too small, missing data, or respondents who do not match the target population, changing an SPSS option will not fix the cause. Discuss with your supervisor whether you can collect more data or limit the model. More responses do not guarantee that every indicator will meet its threshold, so do not promise a specific result in advance.
Distinguishing Scientific Research from Quantitative Research
Scientific research is a broad concept that includes many approaches. Quantitative research uses numerical data, scales, and statistical tests to describe or examine relationships between variables. Qualitative research focuses on spoken, written, or observational data to understand meaning and experience in depth. This article focuses on quantitative research using questionnaires, SPSS, and SmartPLS.
Another common confusion is treating a method and a tool as the same thing. A questionnaire is a data collection tool, SPSS is analysis software, and regression or EFA is an analysis method. You can use the same questionnaire with different analyses depending on the research question and model.
Scientific research is also different from summarising the literature. A literature review provides the basis for choosing variables, scales, and hypotheses. Your study needs to go one step further by collecting data in a defined context and testing the claims with a transparent method.
Common Mistakes
Choosing a Topic That Is Too Broad
“The impact of technology on society” is difficult to turn into a feasible survey model within the time available for a thesis. Limit the industry, location, target group, behaviour, and time period. A narrow topic that can be measured is usually stronger than a large topic without suitable data.
Writing Hypotheses After Seeing the Results
If you choose only relationships with small p-values after running the data, the hypotheses lose their direction. Write the model and hypotheses before the formal analysis, and clearly distinguish planned tests from analyses that arose after examining the data.
Mixing Criteria from Different Programs
SPSS regression, AMOS, and SmartPLS use different output tables and assessment logic. In particular, CB-SEM fit indices should not be transferred into PLS-SEM assessment without a reason. When you change software, change the reporting procedure as well, not only the screenshots.
Showing Output Without Interpretation
A Coefficients screenshot does not tell the reader whether a hypothesis is supported or rejected. In the results chapter, state which variable has an effect, the direction of the effect, the coefficient, the significance level, and the conclusion for the corresponding hypothesis. The discussion section explains why the result is similar to or different from earlier research.
Frequently asked questions
What is scientific research if I only conduct a questionnaire survey?
A questionnaire survey is still scientific research if you have clear questions, an appropriate sample, supported scales, a transparent data collection process, and a consistent analysis method. A large number of responses does not automatically make a study better if the respondents do not match the target population or the items do not measure the intended concept.
Does scientific research have to use SPSS?
No. SPSS is a common choice for Cronbach's Alpha, EFA, correlation, and regression. SmartPLS is suitable for PLS-SEM, AMOS is suitable for CB-SEM, and JASP or R can support many other analyses. Choose the tool based on the model, your ability to use the software, and your supervisor's requirements.
How large must the sample be for scientific research to be acceptable?
There is no single sample size for every project. For regression, the rule of 50 plus 8 times the number of independent variables is commonly cited according to (Tabachnick and Fidell, 2013). For scales, 5 to 10 observations per item is a reference point according to (Hair et al., 2010). You still need to review the sampling design and valid response rate.
What should I do if Cronbach's Alpha does not meet the criterion?
First check reverse-coded items, missing values, coding, and the wording of each item. Then review Corrected Item-Total Correlation and Alpha if Item Deleted. You can delete an item when statistical and theoretical evidence support the decision, but record the reason. If the scale is new and exploratory, Alpha from 0.60 may be used as a reference according to (Hair et al., 2010).
Can I change my hypothesis after seeing the output?
You can conduct an additional analysis if you identify an interesting result, but you must state that it was an exploratory analysis. Do not rewrite the original hypothesis so that an unexpected result appears to have been predicted. Transparent reporting helps the reader distinguish the original design from the interpretation after analysis.
How to write this in your thesis
Open your data file now and check the research question, variable list, and analysis order before running the next table. If you need to run the analysis on your own .sav or .csv file, see the M4 data analysis module.