Common Statistical Reporting Errors in Theses: How to Stop “Lying” with Data (2026 Guide)

You have spent months collecting data. You finally figured out how to run the analysis in SPSS or SmartPLS. You see a table full of numbers. You feel relieved.

So, you do what 90% of postgraduate students do: You take a screenshot (or copy-paste) the entire table into Chapter 4 and hit “Save.”

This is a guaranteed way to get “Major Corrections.”

Reporting statistics is not just about showing the numbers; it is about showing the right numbers in the right format with the right interpretation.

Examiners at universities like UiTM, USM, and UPM are trained to spot “statistical lying”—intentional or accidental. If you report a P-value of 0.000 as “zero error,” or if you claim “Causation” when you only tested “Correlation,” you are technically lying to the reader.

Here are the 6 Most Common Statistical Reporting Errors in Malaysian theses and exactly how to fix them before your Viva.

1. The “$p = 0.000$” Fallacy

This is the most famous mistake in the history of postgraduate research.

The Error:

In SPSS, when a result is highly significant, the output often displays the Sig. (Significance) value as .000.

• Student writes: “The result showed a significance level of $p = 0.000$.”

The Reality:

There is no such thing as “zero” probability in statistics. There is always a tiny chance of error. SPSS just cuts off the number because it ran out of space. It actually means $0.00000012…$

The Fix:

Never write $p = 0.000$.

• Correct APA Format: Write $p < 0.001$.
• Why: This tells the examiner, “The chance of error is less than 0.1%, which is statistically significant.” It looks professional and accurate.

2. The “Kitchen Sink” Table (The Copy-Paste Crime)

The Error:

You copy the entire “Model Summary” or “Coefficients” table from SPSS directly into Microsoft Word.

• Why it fails: SPSS output is ugly. It has vertical lines (forbidden in APA). It contains columns that nobody needs (like “Sum of Squares,” “Mean Square,” or “F-Change” for a simple regression).

The Reality:

Examiners are lazy readers. If you give them a table with 15 columns, they won’t read it. They will assume you don’t know which numbers matter.

The Fix:

Re-type the table. Create a new table in Word with only the columns that matter for your hypothesis:

1. Variable Name
2. Beta ($\beta$) (The strength of the relationship).
3. t-value (The test statistic).
4. p-value (The significance).
5. (Optional) $R^2$ (How much variance is explained).

• Rule: If you don’t discuss a number in the text, delete it from the table.

3. Decimal Diarrhea (False Precision)

The Error:

Reporting numbers with 5 or 6 decimal places.

• Student writes: “The mean satisfaction score was 4.12859.”

The Reality:

This implies a level of precision that doesn’t exist. On a 5-point Likert scale (1 to 5), does “4.12859” really mean anything different from “4.13”? No. It just makes the text hard to read.

The Fix (APA 7th Guidelines):

• P-values: Report to 3 decimal places (e.g., $p = 0.042$).
• Everything else (Means, SD, Beta, t-values): Report to 2 decimal places (e.g., $M = 4.13$, $SD = 0.85$).
• Exception: If you are doing structural engineering or chemistry where microscopic differences matter, follow your faculty guideline. For Social Sciences, 2 decimals is the law.

4. Confusing Correlation with Causation (The “Impact” Trap)

The Error:

Your Research Objective says “To determine the relationship…” (Correlation).

But your Chapter 4 Conclusion says “Therefore, Price impacts Sales.” (Regression/Causation).

The Reality:

• Correlation ($r$): Tells you if two things move together. (When it rains, umbrella sales go up).
• Regression ($\beta$): Tells you if one thing causes the other. (Rain causes umbrella sales).

You cannot use the word “Impact,” “Effect,” or “Influence” unless you ran a Regression Analysis (or Structural Equation Modeling). If you only ran a Pearson Correlation, you can only say “Associated with” or “Linked to.”

The Fix:

Check your verbs.

• Pearson Correlation: Use “Related,” “Associated,” “Linked.”
• Regression / PLS-SEM: Use “Predicts,” “Influences,” “Affects,” “Impacts.”

5. Ignoring Normality (The “Hidden” Assumption)

The Error:

Jumping straight to Hypothesis Testing without checking if the data is “Normal” (Bell Curve).

• Scenario: You use parametric tests (like t-tests or ANOVA) on data that is heavily skewed (e.g., Income data is usually skewed because billionaires ruin the average).

The Reality:

If your data is not normal, your P-values are wrong. You are making false claims.

The Fix:

In Section 4.2 of your thesis, you MUST report Skewness and Kurtosis.

• Acceptable Range: Usually between -1.0 and +1.0 (or -2 to +2 depending on the reference, e.g., George & Mallery, 2010).
• What if it fails? You must justify why you proceeded.
  • Justification: “According to Hair et al. (2019), PLS-SEM is robust against non-normal data.” (This is why SmartPLS is so popular in Malaysia—it saves you from this headache).

6. Confusing Cronbach’s Alpha with Composite Reliability

This is specific to users of SmartPLS / PLS-SEM.

The Error:

Reporting Cronbach’s Alpha as the only measure of reliability because that’s what your senior did.

The Reality:

Cronbach’s Alpha is an “old school” metric (conservative). It often underestimates reliability.

If you are using PLS-SEM, the modern standard is Composite Reliability (CR) or Rho_A.

The Fix:

• If using SPSS: Report Cronbach’s Alpha (> 0.7).
• If using SmartPLS: Report Composite Reliability (> 0.7) AND Average Variance Extracted (AVE > 0.5).
• Pro Tip: If your Cronbach’s Alpha is low (e.g., 0.65) but your CR is high (0.80), report the CR and cite Hair et al. (2017) to save your variable from being deleted.

Summary Checklist: The “Accuracy” Test

Before you submit Chapter 4, scan every page for these red flags:

1. Did I write $p = 0.000$? (Change to $p < 0.001$).
2. Are my tables copy-pasted from SPSS? (Re-type them in Word with horizontal lines only).
3. Do I have 5 decimal places? (Round to 2 decimal places).
4. Did I say “Impact” when I only tested “Correlation”? (Fix the verb).
5. Did I report Normality (Skewness/Kurtosis)? (Add this section before the Regression).

Statistics are the “evidence” in your court case. If the evidence is messy or exaggerated, the judge (examiner) will throw the case out. Keep it clean, precise, and honest.

Are your tables looking messy?

• Get a Statistical Review and Professional Thesis Editing: We check your SPSS/SmartPLS output to ensure you aren’t making fatal interpretation errors.
• APA Formatting Services: We convert raw output into publication-ready, examiner-friendly tables.
• Data Analysis Help: Stuck with “Not Significant” results? We can help you find the reason (moderators/mediators) so you still have a story to tell.