Overview
This data analysis chapter example walks through seven extracts from an illustrative panel-data results section on leverage and profitability among Qatar Stock Exchange firms, with a note on why each one works. A strong chapter states what was done, shows what was found and explains what it means.
Illustrative only. The topic, extracts and table layout are examples written to show the standard. They contain no real data or findings. Use them to understand how to present finance thesis results, not as text to copy.
At a glance
- Name the market, period and panel structure in the first paragraph.
- State how outliers were treated.
- Justify the estimator with a test, not just a name.
- Report sign, significance and economic size together.
- Show your robustness checks and state the limits of the design.
How to Read an Annotated Report
Each extract in this results section sample is followed by a note. As you read, ask three questions: does it link back to the research question, does it justify its choices, and could another researcher repeat it?
The Seven Sections
1. Data and sample
The sample covers non-financial firms listed on the Qatar Stock Exchange from 2010 to 2023, observed annually. The panel is unbalanced because some firms listed or delisted during the period.
Why it works: the market, period and panel structure are named, and the gaps are explained, so an examiner can judge the sample at once.
2. Descriptive statistics
Table 1 reports the mean, standard deviation and range of each variable. Financial ratios were winsorised at the 1st and 99th percentiles to limit the effect of outliers.
Why it works: the treatment of extreme values is stated openly, so the reader knows how the data were prepared.
3. Model specification
Return on assets is regressed on leverage, firm size and sales growth, with firm and year fixed effects and standard errors clustered by firm.
Why it works: variables, controls and error treatment are set out in one sentence, and each choice can be defended in the viva [1][2].
4. Model selection
A Mundlak test with firm-clustered standard errors rejects the random effects assumption, so fixed effects estimates are reported as the main results. A conventional Hausman test without clustering points the same way.
Why it works: the estimator is justified by a test that suits clustered errors [3][4], not just named. Mentioning the conventional test shows awareness of both.
5. Main results
Leverage is negatively and significantly related to return on assets. A one-standard-deviation rise in leverage is associated with a fall in ROA of 0.8 percentage points, which is economically meaningful relative to the sample mean of 4.2% (illustrative figures).
Why it works: sign, significance and size are reported together and linked back to the hypothesis.
Table template: how the main results table should look
| (1) Pooled OLS | (2) Fixed effects | (3) Random effects | |
|---|---|---|---|
| Leverage | β (SE) | β (SE) | β (SE) |
| Firm size | β (SE) | β (SE) | β (SE) |
| Sales growth | β (SE) | β (SE) | β (SE) |
| Firm fixed effects | No | Yes | No |
| Year fixed effects | Yes | Yes | Yes |
| Firms / Observations | N / n | N / n | N / n |
| Within R² | — | value | — |
| Mundlak test (p-value) | value |
Notes: standard errors clustered by firm in parentheses. *, **, *** = significance at 10%, 5%, 1%.
6. Diagnostics and robustness
Tests for serial correlation and cross-sectional dependence supported clustering by firm. The main result holds with an alternative measure of profitability (return on equity) and when 2020 is excluded.
Why it works: checks are reported, not hidden, and each one answers a likely examiner question [1][2].
7. Interpretation and limits
The negative association is consistent with the view that higher debt adds financial pressure in a small market. However, the study covers listed firms only and cannot show causation.
Why it works: the result is tied to theory, a limit is stated, and no causal claim goes beyond what the design supports.
What the Sample Shows
Every extract reuses the core terms of the research question, which keeps one thread running from data to interpretation. The standard to aim for is a clearly described sample, a justified estimator, results reported with their size and meaning, and honest checks [1]. Keeping every step in a script also lets others repeat your work [5]. The notes never add new evidence. They only explain the choices. When you review your own draft statistical analysis report, compare each section with these notes.
FAQ
- What should a data analysis chapter include?
Sample and data, descriptive statistics, model specification, model selection, main results, robustness checks, and interpretation with limits.
- How do I report fixed effects results?
In a table alongside pooled OLS and random effects, with the number of firms and observations, the fixed effects included and the clustering level.
- Can I copy this structure?
Yes, the structure of this data analysis chapter example. The text is illustrative and must be replaced with your own analysis and wording.From Core Results to Newer Methods
Once your core results are sound, you can extend them with machine learning, causal designs or fully reproducible workflows. Read Recent Trends in Finance Data Analysis: Machine Learning, Causal Inference and Reproducible Stata/R Workflows for Qatar Doctorates.
For finance doctoral scholars in Qatar
A strong results chapter needs the right model, tested assumptions and clear reporting. PhD Assistance’s econometrics team reviews panel analyses and results chapters, from model selection to interpretation. Our finance data analysis and econometrics support guides Qatar doctoral scholars through model selection, Stata and R coding and results interpretation.
Related reading: How to run your first panel regression in Stata · Reporting econometric results in finance
