What Is Information Gathering? A Process for Your Thesis
What Is Information Gathering
Information gathering is the process of identifying, obtaining, and recording the data you need to answer a research question or test a hypothesis. In a quantitative thesis, that information is usually converted into data columns in a .sav or .csv file: one row represents one respondent, and one column represents an item or a categorical variable.
Start with your research model rather than opening Google Forms and thinking about the questions afterward. If your hypothesis tests the effect of service quality on satisfaction, your questionnaire must collect items that represent both constructs, along with the sample description information you need.
There are two main sources. You collect primary data directly from respondents using a questionnaire. Secondary data existed before you began the study, such as reports, databases, public statistics, or business records that you are permitted to use. A thesis can use one source or combine both, but every number needs a clear origin and recording process.
For a Likert survey, each item is coded as a numerical value, usually from 1 to 5 or from 1 to 7. The Likert scale originated from Likert's attitude measurement technique (Likert, 1932). Record the exact coding convention in your data codebook, such as 1 meaning strongly disagree and 5 meaning strongly agree.
Why Information Gathering Matters in Quantitative Research
Information gathering determines whether your data actually answer your research topic. A file with 500 rows still provides unsuitable evidence if the respondents are not the target population. A moderate sample with consistent sampling criteria, questions, and recording procedures is easier to explain when you defend your thesis.
Four decisions need to be settled before you distribute the questionnaire. First, decide who you will collect information from. Second, specify which items measure each construct. Third, decide when and through which channel you will collect the responses. Fourth, define the criteria for keeping or removing a response.
These decisions also affect the analyses that follow. Responses with too much missing information reduce the number of observations available for Cronbach's Alpha, EFA, or regression. Inconsistent coding can cause SPSS to read a variable incorrectly. An unreversed item that has not been recoded can produce low correlations between items even when the scale content is reasonable.
You can review the basic concept in what is a scale before writing your questionnaire. If your model includes a mediator or moderator, identify them at the beginning because each variable needs its own scale and items. Adding items after data collection has started often leaves one group of variables with incomplete data.
Information gathering also helps you control bias. A respondent may select the same response for the entire questionnaire, finish too quickly, submit more than once, or fall outside the survey population. Do not delete a record automatically just because it has a high score. Set the criteria first, check the data against them, and keep a processing log.
How Much Information Is Enough
Being sufficient does not only mean having enough rows. A dataset may meet the numerical target but still fail on respondent eligibility, completeness, response quality, or traceability. Use the thresholds below as planning references, not as permission to replace a sampling design.
| Check | Reference point or rule | Canonical source |
|---|---|---|
| Observations per item | Approximately 5 to 10 observations for each item | (Hair et al., 2010) |
| Regression sample size | At least 50 + 8m, where m is the number of independent variables | (Tabachnick and Fidell, 2013) |
| KMO for EFA | KMO of 0.5 or above | (Kaiser, 1974) |
| Bartlett's Test | p-value below 0.05 | (Kaiser, 1974) |
| Corrected Item-Total Correlation | 0.3 or above | (Nunnally and Bernstein, 1994) |
| Cronbach's Alpha | 0.7 or above, with 0.6 potentially acceptable for an exploratory scale | (Nunnally, 1978); (Hair et al., 2010) |
| Total variance explained in EFA | 50% or above | (Hair et al., 2010) |
| Eigenvalue when deciding the number of factors | Above 1 | (Kaiser, 1960) |
Suppose your questionnaire has 30 items. The rule of 5 to 10 observations per item suggests approximately 150 to 300 usable responses. If you run a regression with 6 independent variables, the formula 50 + 8m produces a minimum of 98 observations under the cited rule. These calculations serve different purposes, so state clearly which basis you use in the methods chapter.
A calculated sample size does not mean you should stop collecting as soon as you reach that number. Allow for invalid, incomplete, or ineligible responses. Your reserve rate should reflect your collection channel and pilot experience. Do not write an exact percentage without a basis.
If the population is finite and you have a reliable population-size list, you can present the finite-population sample-size formula from Yamane (1967). If the population size is not known precisely, do not insert a formula into your thesis simply because you found it online. Explain the population, how respondents were selected, and the final number of valid responses.
How to Read Information in Your Data and Output
After closing the survey, inspect the data file before running any analysis. In SPSS, open Variable View to check variable names, variable labels, data types, coded values, and missing values. In Data View, inspect several rows at the beginning and end of the file, as well as unusual values. A Likert variable entered as a string will not run correctly in many analysis dialogs.
The table below is illustrative output, not the result of a real study. It simulates a data-check summary after 220 responses were collected, with 198 responses retained under the criteria set in advance.
| Check item | Column name or information in the file | Illustrative output | How to read it |
|---|---|---|---|
| Total responses received | N_received | 220 | Number of records downloaded from the form |
| Responses removed for substantial missing data | N_incomplete | 12 | Compare with the stated removal criteria |
| Duplicate responses | N_duplicate | 4 | Check response codes, timestamps, or duplicate patterns |
| Ineligible responses | N_ineligible | 6 | Based on the screening questions |
| Valid sample size | N_valid | 198 | Number of rows entered into SPSS |
| Number of Likert items | n_items | 30 | Used to check the observations-per-item ratio |
| Valid data rate | N_valid / N_received | 90,0% | Describes the process, not the only measure of quality |
To check missing values, go to Analyze > Descriptive Statistics > Frequencies, move the variables into the analysis box, and inspect the frequency table. For quantitative variables, Analyze > Descriptive Statistics > Descriptives helps identify values outside the coding range, such as a Likert variable with a value of 8 when the codebook permits only values from 1 to 5.
To run a reliability analysis, go to Analyze > Scale > Reliability Analysis. Select the items belonging to the same construct, open Statistics, and check Item, Scale, and Scale if item deleted. In the Reliability Statistics table, read Cronbach's Alpha and N of Items. In Item-Total Statistics, pay attention to Corrected Item-Total Correlation and the Cronbach's Alpha if Item Deleted column.
For example, if one row has Corrected Item-Total Correlation = 0.218, it is a signal to review the item against the 0.3 reference point from (Nunnally and Bernstein, 1994). You still need to read the item wording, check whether it is reverse-coded, and inspect what happens if you remove it. Do not delete it immediately just to increase Alpha.
Once the data meet the preliminary conditions, run EFA or regression according to your model. For EFA, the tables you usually see are KMO and Bartlett's Test, Total Variance Explained, and Rotated Component Matrix. For regression, the Coefficients table contains B, Beta, t, and Sig., with Sig. commonly read as the p-value when making a hypothesis decision.
What to Do When Information Gathering Falls Short
Work from problems that can be verified toward issues that require redesign. First, save an untouched original file. Create a working copy and record the codes of removed responses, the reason for removal, and the processing date. This gives you an answer when your supervisor asks why the number of received responses differs from the number entered into the analysis.
Step 1: Check the Screening Conditions
Check whether respondents belong to the correct location, age group, experience group, or relationship with the target population. If a respondent answered “no” to a screening question but continued filling out the questionnaire, remove the response according to the rule set in advance. Do not change the criteria simply to reach the desired sample size.
Step 2: Handle Incomplete Responses
Distinguish between a few blank cells and a questionnaire that was abandoned. If missing data occur in a key variable, you may remove the entire response according to the stated criterion. Mean imputation should be considered only when it fits the design and can be explained clearly, because it creates data that the respondent did not directly provide.
Step 3: Check for Inattentive Responses
Possible signs include an unusual completion time, selecting one column for an entire scale, or repeating the same response pattern across several submissions. These signs support a review, but they are not sufficient on their own. Combine several indicators and record the rule before removing a response.
Step 4: Collect More Responses if the Sample Is Still Too Small
If the valid sample is not large enough, collect more responses from the correct population while keeping the questionnaire, coding, and criteria unchanged. If the problem is unclear wording or a scale that measures the wrong construct, collecting more responses with the same error will not solve the data problem. Return to the questionnaire, record the changes, and discuss the next collection round with your supervisor before opening it.
If your data contain extreme values, check outliers as a separate issue. An outlier is not automatically a fake response. Deleting a row simply because it makes the results look better is difficult to defend.
Distinguishing Information Gathering from Secondary Data Collection
Information gathering is the broader concept. It includes obtaining primary data and finding secondary data. Primary data collection starts with a questionnaire, a sample design, and new responses. Secondary data collection starts with an existing source, followed by checks of usage rights, timing, measurement units, and fit with the research model.
| Criterion | Primary data | Secondary data |
|---|---|---|
| Data creator | You organize the direct collection | An agency, business, or previous study |
| Common tools | Questionnaire, online form | Report, database, record, statistical yearbook |
| Main checks | Population, questions, responses, missing data | Source, year, variable definition, measurement method |
| Main risk | Sampling bias, inattentive responses | Conceptual mismatch or an untraceable source |
| How to present it | Questionnaire distribution and cleaning process | Source name, date, scope, and extraction method |
Secondary data are not automatically more reliable than primary data. You need to know how the variable was measured, whether its definition changed across years, and whether its unit is compatible with your analysis. If you use a report to provide context but use a questionnaire to test hypotheses, separate those two roles clearly in the methods chapter.
Common Mistakes
The first mistake is collecting data before defining the variables. You then have many responses but lack the variables needed to test your hypotheses. Create a table linking each hypothesis, construct, item, questionnaire item, and variable code before distributing the questionnaire.
The second mistake is using several names for the same variable. For example, HL1, HL_1, and HaiLong1 appear in different files. Choose a short naming rule without diacritics or spaces, and keep one codebook.
The third mistake is failing to save versions. Name files by date, such as raw_2026-09-08.csv and clean_v01.sav, then record every removal of an item or response. In SPSS, you can use Transform > Recode into Different Variables to preserve the original variable and create a reverse-coded variable instead of overwriting the initial data.
The fourth mistake is confusing the number of responses received with the valid sample size. In Chapter 3, distinguish between questionnaires distributed, responses received, removed responses, and final observations. If you use percentages, calculate them from a clearly stated denominator.
The final mistake is treating a tool as a solution to a methodological problem. Google Forms works for creating a basic form, while fillform.info is DoThesis's own form tool for collecting responses to Vietnamese questionnaires. Whichever tool you use, check the population, removal criteria, and exported file. If you need to analyze a relationship involving a moderator, identify the variable before collecting responses and review moderator or moderating.
Frequently asked questions
What is information gathering in a thesis?
It is the full process of identifying the information you need, selecting sources, designing the instrument, recording responses, and checking the data before analysis. In a quantitative thesis, the final result is usually a structured data file that can be analyzed in SPSS or SmartPLS.
What is the difference between primary and secondary data collection?
You collect primary data directly from the research population, usually through a questionnaire. Secondary data already existed, so you need to check the source, date, variable definition, and usage rights before including it in the study.
How many responses are enough for information gathering?
There is no single number that applies to every topic. Base your decision on the number of items, number of independent variables, analysis method, and sample design. You can refer to the ratio of 5 to 10 observations per item from (Hair et al., 2010), or the formula 50 + 8m for regression from (Tabachnick and Fidell, 2013).
What should I do with missing data?
First, identify which response contains the missing cell and how much information is missing. Keep the original file, apply the criteria set in advance, record removed responses, and consider a missing-data treatment only when you can explain it in the methods chapter.
Should I delete a response that uses the same rating for the whole questionnaire?
This is a signal to check, but it is not enough to conclude that the response is invalid. Also examine completion time, screening answers, repeated patterns, and response content. Set the removal criteria in advance or explain them transparently rather than basing the decision on whether the results look better.
Should I use Google Forms or another tool for information gathering?
Both Google Forms and fillform.info can export data for processing, but you need to check column formats, coded values, and duplicate responses after downloading the file. The tool only supports collection. Research quality depends on questionnaire design and sample control.
Open your data file now, create an untouched original copy, build the variable codebook, and compare each column with your hypotheses before collecting more responses or cleaning the file. If you need to run analyses on your own .sav or .csv file, DoThesis M4 analysis performs this step with SPSS and SmartPLS.