What Is Quantitative Research? Quantitative vs. Qualitative
What Is Quantitative Research
Quantitative research uses data that can be represented numerically to measure variables, test hypotheses, and assess relationships between variables. You usually collect data with a questionnaire containing fixed response options, code the responses as numbers, and then analyse them with SPSS, SmartPLS, AMOS, JASP, or R.
For example, your study may test whether service quality affects customer satisfaction. Service quality and satisfaction are concepts that need to be measured with multiple observed variables. Respondents may rate each item on a 5-point Likert scale, from 1 for strongly disagree to 5 for strongly agree, following the attitude measurement approach introduced by (Likert, 1932).
A quantitative research process usually looks like this:
Research question → research model and hypotheses → questionnaire → numerical data → scale assessment → model assessment → conclusion.
The key point is that quantitative research involves more than entering data into SPSS and reading the p-value. You need to define the concept to be measured, select or adapt a scale, collect enough observations, clean the .sav or .csv file, choose appropriate analyses, and interpret the results against the hypotheses. If you are just starting, you can also read what is scientific research to place the quantitative part within the full research process.
The Role of Quantitative Research in Research
In a thesis, quantitative research helps you answer questions that can be converted into variables and numbers. For example, you can examine how strongly factors affect purchase intention, whether two groups differ in their mean scores, or what proportion of people use a service. The clearer the question, the less likely you are to choose the wrong data or statistical test.
Quantitative research is usually appropriate when your model contains relatively clear concepts that can be represented by indicators and when you need to test relationships such as “X affects Y.” With a model containing latent variables, you can assess the reliability and validity of the scale first, then evaluate the relationships between variables in the model. Assessing the measurement part before the structural part also follows the two-step logic presented by (Anderson and Gerbing, 1988).
The main decisions that quantitative data can support include:
- Describing the sample: Which age, gender, occupation, or usage-frequency groups do the respondents belong to.
- Assessing the scale: Whether the items measure the same concept.
- Testing relationships: Whether an independent variable is related to or affects a dependent variable as hypothesised.
- Assessing explanatory power: How much variation in the dependent variable the model explains.
- Comparing groups: Whether groups differ in their mean scores, when the research design includes this question.
You should distinguish “quantitative” from “a large dataset.” A survey with several hundred rows may still fail to answer the research question if the variables are measured incorrectly or the sample is unsuitable. Conversely, a small dataset can still be useful for exploratory analysis when it has been designed properly and you state the limitations of the results clearly.
You can also read what is qualitative research to see how the data type, objective, and reasoning differ. If you are defining the scope of your topic, the pages on research fields and research subjects can also help you check who you are studying and in what context.
What Counts as Acceptable in Quantitative Research
There is no single number that determines whether quantitative research is acceptable. Sample size, scale reliability, EFA, regression, and PLS-SEM each use different criteria. You also need to consider the research objective, scale type, and model size before drawing a conclusion. The table below summarises common thresholds and gives you a source to cite in your thesis.
| Check | Common threshold | Source in the methods literature |
|---|---|---|
| Cronbach's Alpha | 0.7 or above | (Nunnally, 1978) |
| Cronbach's Alpha for a new or exploratory scale | May be from 0.6 | (Hair et al., 2010) |
| Corrected Item-Total Correlation | 0.3 or above | (Nunnally and Bernstein, 1994) |
| KMO | 0.5 or above | (Kaiser, 1974) |
| Bartlett's Test | p-value below 0.05 | (Kaiser, 1974) |
| Eigenvalue in EFA | Above 1 | (Kaiser, 1960) |
| Total Variance Explained | 50% or above | (Hair et al., 2010) |
| Factor loading in EFA | 0.5 or above | (Hair et al., 2010) |
| Observations per item | Approximately 5 to 10 | (Hair et al., 2010) |
| VIF in a PLS-SEM model | Below 5 | (Hair et al., 2019) |
| R² in PLS-SEM | 0.75 substantial, 0.50 moderate, 0.25 weak | (Hair et al., 2011) |
| HTMT | Below 0.85, or below 0.90 for closely related concepts | (Henseler et al., 2015) |
These thresholds are reference points, not automatic reasons to delete an item. For example, a high Alpha can occur when items repeat the same content. A low factor loading needs to be considered alongside the item's theoretical meaning, cross-loading, and the effect of deleting it. When you use SmartPLS, keep the PLS-SEM indicators together, including outer loading, CR, AVE, HTMT, VIF, R², and Q². CFI, TLI, and RMSEA belong to CB-SEM. They are discussed for CB-SEM by (Hu and Bentler, 1999), so do not place them in the SmartPLS assessment section.
How to Read Quantitative Results in Output
You cannot read one isolated number and use it to conclude that the entire study works. Move from descriptive statistics, reliability, and scale validity to the model and hypotheses. Table names depend on the software. In SPSS, common tables include Reliability Statistics, Item-Total Statistics, KMO and Bartlett's Test, Total Variance Explained, Rotated Component Matrix, and Coefficients. In SmartPLS, you will see Outer Loadings, Discriminant Validity, Path Coefficients, and R Square.
The table below contains illustrative output, not the result of a real study. It simulates part of SPSS output for a model with the independent variable CLDV and the dependent variable HL.
| SPSS output table | Indicator or row | Illustrative data | How to read it |
|---|---|---|---|
| Reliability Statistics | Cronbach's Alpha, CLDV | 0.842 | The internal consistency of the CLDV scale reaches the reference level |
| Item-Total Statistics | Corrected Item-Total Correlation, CLDV3 | 0.517 | CLDV3 exceeds the reference threshold |
| KMO and Bartlett's Test | KMO | 0.781 | The data are suitable for considering EFA under the selected threshold |
| KMO and Bartlett's Test | Bartlett's Test Sig. | 0.000 | You can reject the hypothesis that the observed correlations equal 0 at the 5% level |
| Total Variance Explained | Cumulative % | 61.438 | The retained factors explain the reported share of variance |
| Rotated Component Matrix | Factor loading, CLDV2 | 0.736 | CLDV2 has a loading on the corresponding factor |
| Coefficients | Standardized Beta, CLDV | 0.428 | The standardised effect coefficient in the illustrative model |
| Coefficients | Sig., CLDV | 0.003 | The p-value is below 0.05 in the example, so compare it with the hypothesis |
When reading the Coefficients table, first identify the variable being examined in the Model column or in the row name. Then check the sign of Standardized Coefficients Beta to determine whether the relationship is positive or negative, check Sig. to assess the statistical evidence, and record VIF to examine multicollinearity. The conclusion “the hypothesis is supported” is appropriate only when the sign, p-value, model, and hypothesised direction are consistent.
If you use SmartPLS, keep the two parts separate. Outer Loadings, CR, AVE, and HTMT belong to the measurement model. Path Coefficients, p-value, confidence interval, R², f², and Q² belong to the structural model. In PLS-SEM, bootstrapping is commonly run with 5,000 bootstrap subsamples following the guidance of (Hair et al., 2022). The number in your output is your research result. The table above is only for practising how to read the columns.
What to Do When the Quantitative Results Do Not Meet the Threshold
If an indicator does not meet the threshold, work through the issue in an order that you can explain, rather than deleting variables until the table looks better. Start by checking the original data: Was a variable coded incorrectly, how were missing values defined, was a reverse-coded item recoded, and did any column name change when the data were imported from Excel?
Next, inspect each item in Item-Total Statistics, Rotated Component Matrix, or Outer Loadings. An item with a low item-total correlation or low loading may be problematic, but read the item's content and theoretical basis before deleting it. If you delete an item, save a separate data file and output file for each run, and record which item was deleted and why.
If EFA does not produce the number of factors specified in the model, check how the variables were entered, the extraction method, the rotation method, and whether some concepts have content that is too similar. If the questionnaire includes reverse-coded items, check the coding again before concluding that the scale is weak. When the sample is too small or the responses are too homogeneous, the software cannot repair a design problem.
There are situations where you should stop adjusting the output. For example, you may delete an item only because you want Alpha to increase, even though that item is an important part of the concept. The committee may ask why the original scale contained that item while the results chapter no longer does. A safer approach is to report the criteria, the number of assessment rounds, the deleted items, and the effect on the model.
How to write this in your thesis
[State the criterion used]. The results in the [...] table show [the relevant indicator and value]. Because [interpretation linked to the criterion], [the item, scale, or relationship] was [retained or removed] before the next analysis.
Quantitative and Qualitative Research Compared
Quantitative and qualitative research differ first in their objectives and data types. Quantitative research uses numerical data to measure, describe, test, or estimate relationships. Qualitative research usually focuses on meaning, experience, and how participants interpret a phenomenon. You can read the article on scientific research to place both approaches in a broader context.
| Criterion | Quantitative | Qualitative |
|---|---|---|
| Common questions | How many, to what extent, and how does X affect Y | Why, how, and how is an experience understood |
| Data | Scores, frequencies, proportions, and coded variables | Participants' expressed content and meanings |
| Common tools | Questionnaires, SPSS, SmartPLS, AMOS, JASP, R | Interviews, observation, documents, and text-processing software |
| Results | Coefficients, p-values, confidence intervals, R², and distributions | Themes, interpretations, and contextual descriptions |
| Reasoning for conclusions | Comparing hypotheses with numerical evidence | Clarifying how a phenomenon is described or explained |
A topic can begin with a quantitative question when you already have a sufficiently clear model and scale. If the question is still too broad, such as wanting to know only “how a phenomenon occurs,” narrow it into variables, subjects, context, and hypotheses before opening SPSS. Quantitative research also does not mean that every result can be generalised to the entire population. Generalisability also depends on the sampling method, sample size, questionnaire quality, and research scope.
Common Mistakes
Calling every survey quantitative research. A questionnaire with open responses or data that were not designed to measure variables does not automatically constitute a complete quantitative study. You need to identify the variables, coding method, and analysis you will use.
Choosing a test based on the topic title. The word “effect” in a title is not enough to determine whether you should use regression or PLS-SEM. Look at the number of variables, indicators, scale type, sample size, and hypothesised model.
Taking thresholds from an unsourced post. When writing Chapter 3 or Chapter 4, report both the criterion and its source. Alpha, KMO, factor loading, HTMT, and R² do not share one common threshold.
Treating the p-value as the entire result. The p-value helps you assess statistical evidence under the selected convention, but it does not replace the effect coefficient, confidence interval, effect size, or practical meaning.
Deleting items until the output looks good. Each deletion changes the scale and may change the hypothesis or the meaning of the concept. Keep a processing history so that you can explain the decisions to your supervisor.
Mixing SPSS and SmartPLS indicators. EFA and linear regression in SPSS are read differently from the measurement and structural models in SmartPLS. State the software, version, analysis procedure, and corresponding output table clearly.
Frequently asked questions
What is quantitative research in science?
Quantitative research uses numerical data to measure variables and test a research question or hypothesis. The data usually come from a questionnaire with coded response options, followed by analysis in SPSS, SmartPLS, or another statistical program.
Do you have to use SPSS for quantitative research?
No. SPSS is widely used for descriptive statistics, Cronbach's Alpha, EFA, regression, and many other tests. SmartPLS is suitable for PLS-SEM, while AMOS, JASP, and R each have their own strengths. Choose based on the model, data type, supervisor's requirements, and your ability to explain the results at the defence.
What sample size is acceptable for quantitative research?
There is no single number that applies to every study. For scales, one commonly cited guideline is approximately 5 to 10 observations per item (Hair et al., 2010). For regression, the reference formula is n ≥ 50 + 8m, where m is the number of independent variables, according to (Tabachnick and Fidell, 2013). You still need to consider the number of variables, analysis method, and sampling approach.
What is the difference between quantitative and qualitative research?
Quantitative research answers questions about levels, proportions, relationships, or effects with numerical data. Qualitative research focuses on meaning and how a phenomenon is interpreted. When your topic already has variables, a scale, and hypotheses to test, the quantitative approach is usually more suitable.
What should you do if quantitative data contain errors?
Check the original file, variable coding, missing values, reverse-coded items, outliers, and inconsistent responses before running the analysis again. Then inspect each output table, save each processing version, and record why an item was deleted or retained. If the problem comes from the questionnaire or sampling method, software cannot fully correct it.
Open your data file, write down the dependent variable, independent variables, number of items, and planned analysis, then compare each output table with a sourced criterion. If you need to run the analysis on your own .sav or .csv file and track each processing round, you can use DoThesis M4 quantitative data analysis.