Structural model before measurement model
Testing hypotheses in SEM before the CFA shows acceptable fit, reliability and discriminant validity. Examiners ask about the measurement model first.
PhD statistics help · South Africa
PhD statistics help that holds up under examination
Specialist support for doctoral candidates whose analysis carries the thesis: research design, measurement, modelling, robustness and interpretation. Every engagement is led by Dr Michael Pietersen (PhD, University of Pretoria).
- ✓ Risk Reviews from R4,000
- ✓ Fixed scope before work starts
- ✓ Reproducible code with every result
- ✓ No ghostwriting
Supporting candidates from leading institutions: WBS, GIBS, UCT GSB, Stellenbosch, Henley, Milpark, UNISA SBL
Why candidates get in touch
Where doctoral analysis usually gets challenged
A doctoral thesis is examined on whether its conclusions follow from the evidence. These are the methodological weak points we see most often, and the ones we help you resolve before an examiner finds them.
Endogeneity in panel data
Lagged dependent variables and reverse causality bias fixed-effects estimates. Dynamic panel GMM can help, but only with valid instruments and the right diagnostics.
Too many instruments
GMM results that look strong because the instrument count approaches the number of groups, which weakens the Hansen test. We report and control the instrument count.
Causal claims from cross-sectional data
Mediation tested on a single survey wave can show a theoretical pathway, not prove one. The wording of your results chapter matters as much as the model.
Robustness as a checkbox
Generic sensitivity tests that do not address the threats specific to your design. Robustness should answer the question an examiner is most likely to ask.
Results that cannot be reproduced
Point-and-click output that cannot be re-run when your supervisor asks for one change. Scripted analysis makes every revision quick and traceable.
How we help
Support at every stage of your PhD
The earlier we are involved, the cheaper problems are to fix. You can start at any stage.
-
1
Proposal and design
Research questions mapped to testable models, sample size and power, and a defensible analysis plan.
-
2
Instruments and data
Questionnaire and scale review, data cleaning, missing-data strategy and variable construction.
-
3
Modelling
CFA and SEM, panel models and GMM, time series, fsQCA or multilevel models, with full diagnostics.
-
4
Robustness and interpretation
Context-specific robustness checks and plain-English notes on what each result does and does not show.
-
5
Supervisor feedback and examination
Revisions, examiner reports and corrections, plus preparation for questions on your methods.
Methods we support
Measurement and SEM
- Confirmatory factor analysis (CFA)
- Covariance-based SEM
- PLS-SEM
- Measurement invariance and multi-group SEM
- Mediation and moderation
Econometrics
- Fixed and random effects
- Dynamic panel GMM (difference and system)
- ARDL and cointegration
- GARCH volatility models
- Instrumental variables
Regression and beyond
- Linear, logistic and count models
- Multilevel and hierarchical models
- Fuzzy-set QCA (fsQCA)
- Missing data and multiple imputation
- Applied machine learning
Software
- R (lavaan, plm, SEMinR)
- Python
- SPSS and AMOS conventions
- Stata conventions
- SmartPLS-style reporting
The standard of work
What you actually receive
Not just output, but results you can explain and defend. Below is an excerpt in the format we deliver: a report-ready table, the diagnostics an examiner will check, and an interpretation note.
Example uses a synthetic topic and simulated data. It is not client work.
Example table
Board independence and firm performance: two-step system GMM
| Dependent variable: ROA | Coef. | SE |
|---|---|---|
| ROA (t−1) | 0.412*** | (0.087) |
| Board independence | 0.031** | (0.014) |
| Firm size (log assets) | 0.006 | (0.004) |
| Leverage | −0.058*** | (0.021) |
| Observations / firms | 1,840 / 230 | |
| Instruments | 38 | |
| AR(1) / AR(2), p-value | 0.002 / 0.41 | |
| Hansen J / Difference-in-Hansen, p-value | 0.27 / 0.44 | |
Windmeijer-corrected standard errors. Year effects included. *** p < .01, ** p < .05.
Interpretation note
Board independence is positively associated with ROA (b = 0.031, p < .05) after accounting for performance persistence. Because ROA is persistent, the implied long-run association is larger: 0.031 / (1 − 0.412) ≈ 0.053.
Diagnostics, in plain English
AR(2) is not significant, so there is no evidence of second-order serial correlation, which supports using deeper lags as instruments. Hansen and Difference-in-Hansen do not reject instrument validity, and 38 instruments for 230 firms keeps the count well below the number of groups.
Anticipated examiner question
“Why system rather than difference GMM?” Answer prepared: ROA is highly persistent, so lagged levels are weak instruments for first differences; system GMM adds level equations and the Difference-in-Hansen test supports the extra moment conditions.
Recent engagements (anonymised)
-
Thesis by publication
Development of a composite measure, fsQCA sufficiency analysis, interaction regressions and robustness replications across the papers.
-
Doctoral survey study
Methodology review, data preparation and a regression + fsQCA analysis plan covering three research objectives for a sample of several hundred respondents.
-
Listed-company panel study
A multi-phase analysis pipeline for a disclosure study using firm-level data from two stock exchanges.
Who does the work
Dr Michael Pietersen
Every engagement is led personally by Dr Pietersen, a quantitative researcher who has been through doctoral examination himself. Your analysis is never passed on to a junior analyst.
View publications on Google Scholar-
PhD, University of Pretoria
Doctorate completed in 2022, with a strong quantitative focus. MPhil awarded cum laude.
-
Peer-reviewed researcher
Four peer-reviewed publications, two of them in ABDC A-rated journals.
-
Advanced methods
SEM/CFA, PLS-SEM, panel data and GMM, ARDL and GARCH, fsQCA, and applied machine learning.
-
Reproducible by default
Analysis in documented R or Python code, so every number in your thesis can be traced and re-run.
Client feedback
What researchers say about working with us
★★★★★
“[The support I found most useful was] the structured, continuous engagement to work through each milestone, and the explanations of why certain methods do or don't work. I learnt, for example, how the GMM model and imputation work. I found StatForge to be highly thorough and rigorous in first developing a strategy for the testing methods.”
Anonymous
Postgraduate researcher (name withheld at client's request)
★★★★★
“I appreciated the detailed and in-depth engagement [from Michael]. [Michael] didn't just run the statistical analysis for me, [he] handled everything with such care that I had complete confidence in the results. What really stood out for me though was how [he] caught a few critical weaknesses in my survey. Not only did [he] flag the issues, but [he] also identified possible solutions that I would've never considered. I would recommend [him] without hesitation to anyone doing high-level postgraduate research.”
Aaron
Master of Business Administration, SBS
★★★★★
“I really appreciated the guidance [Michael] provided on the steps I needed to take to ensure that I present and communicate my research and results properly. No amount of reading prepared me for the complexity of what I had to do. I appreciated the assistance in showing me how to set everything up for success.”
“At proposal stage, if you are in doubt about quantitative analysis or mixed-methods, Michael is able to guide you to make the right decision. The research methodology you choose will support a high-performance dissertation and may have a meaningful impact on your future career.”
Deborah
Master’s in Organisational Psychology, UNICAF
Transparent pricing
How much does PhD statistics help cost?
Start with a Risk Review. Larger doctoral projects are quoted in phases, so you only commit to the next stage once the previous one is clear.
Research Design & Analysis Risk Review
R4,000–R8,500
A fixed-fee senior review before you invest in full analysis.
- ✓Pre-read of your proposal and data description
- ✓60–90 minute structured call
- ✓Fatal flaws, model choice and identification issues
- ✓Written memo with risks, fixes and next steps
Analysis Execution
From R9,500
Clean, correct results with deliverables you can write up confidently.
- ✓Data quality review and cleaning
- ✓Assumptions and diagnostics
- ✓Core models for your research questions
- ✓Report-ready tables and reproducible code
Built for doctoral complexity
Research Pro
From R20,000
For complex models and repeated supervisor or examiner feedback.
- ✓SEM/CFA with full validity reporting
- ✓Panel methods and GMM where appropriate
- ✓Robustness designed around your context
- ✓Walkthrough session and extended revision support
Already have results and need supervisor-requested changes? Extensions & Revisions start at R3,500.
How it works
-
1
Free 15-minute scoping call
You share your proposal or chapter. We discuss your study and where you are stuck.
-
2
Fixed scope and quote
You receive the recommended scope, deliverables, timeline and price before any work starts.
-
3
Analysis and validation
Models, diagnostics and robustness, delivered in clear stages with explanations.
-
4
Handover and revisions
Tables, code, data and interpretation notes, plus support through supervisor feedback.
Ethical support, clearly defined
We do
- ✓ Advise on design and methods
- ✓ Prepare data and run analyses
- ✓ Deliver tables, figures and code
- ✓ Explain results so you can defend them
We don't
- ✗ Write chapters or ghostwrite
- ✗ Fabricate or manipulate data
- ✗ Make your interpretive decisions
- ✗ Submit work on your behalf
Your data and drafts are treated as confidential and used only for your project.
Request a quote for your PhD analysis
Tell us about your study. You'll receive a recommended scope, deliverables, timeline and price.
PhD statistics help: frequently asked questions
Can you help with a PhD by publication?
Yes. For a thesis by publication we keep the measurement, analysis code and reporting conventions consistent across papers, so each article stands on its own and the thesis still reads as one coherent body of work.
Is using a statistician for my PhD ethical?
Yes, provided the thesis remains your own work. We provide methodological advice, analysis, reproducible code, tables and plain-English explanations of the results. You write the thesis and make the final interpretive decisions. We do not write chapters or ghostwrite. Seeking specialist statistical advice is common practice in doctoral research; check your institution's policy and keep your supervisor informed.
How much does PhD statistics help cost?
A Research Design & Analysis Risk Review costs R4,000–R8,500. Analysis Execution starts at R9,500 and Research Pro, for complex models and repeated supervisor feedback cycles, starts at R20,000. Larger doctoral projects are usually scoped and quoted in phases, so you only commit to the next stage once the previous one is clear.
Can you work in the software my supervisor prefers?
Yes. Most work is done in R or Python because it is fully reproducible, and outputs are aligned with SPSS, AMOS, Stata or SmartPLS conventions where your supervisor or programme expects them.
Can you help after I receive examiner reports or corrections?
Yes. We review examiner comments on the methods and results, re-run or extend the analysis where needed, and give you a clear account of what changed and why, so you can respond to each point with confidence.
How long does doctoral analysis support take?
A Risk Review is one structured session plus a written memo, scheduled once we have your proposal and data description. Full analysis typically takes two to four weeks or more, depending on data readiness and model complexity. Doctoral projects often run over several feedback cycles with your supervisor.
Do you only work with candidates in Gauteng?
No. All work is done remotely, with candidates at universities across South Africa. Calls take place online and deliverables are shared electronically.
Working on a Master's dissertation or MBA research report? See Master's and MBA statistics help →