RESEARCH SUPPORT
Advanced Stata Analysis for Research
Treatment Effects, Assumptions Stated
Treatment and policy effects estimated in Stata with the estimator that suits your design, overlap and sensitivity examined, and each estimate reported next to the assumptions it depends on, so that a referee can judge how far it can be trusted.
No Commitment Required
Strictly Confidential
In-House PhD-Level Statisticians
Fixed Project Fee
Same-Day Response
What Does Advanced Stata Analysis Involve?
Advanced Stata analysis, for observational and quasi-experimental research, estimates the effect of a program, treatment or exposure on an outcome when allocation was not randomized. Stata treatment effects estimators such as regression adjustment, inverse-probability weighting and doubly robust combinations each rest on assumptions about what drives allocation, and the Stata causal inference commands for instruments and policy changes rest on others. The estimator is matched to your design, fitted on your data and reported beside its identifying assumptions by an in-house PhD-level statistician at OnlineSPSS, a specialist statistical consultancy for academic, clinical and professional research. The substantive claim the results support remains yours to make.
WHEN TO ENGAGE
Research Questions That Ask What a Program Changed
Effect questions are usually put to data collected for another purpose, so the first task is deciding which comparison those data can honestly support.
Allocation Was Not Random Participants chose or were selected into a program, so the groups differ before it starts, and the estimator has to adjust for the factors behind selection that the data actually contain.
A Policy Changed for Some Units A rule, price or law changed for one set of units at a known date, and the design compares changes over time with a comparison group whose trend is a credible stand-in for the treated group.
An Instrument Is Proposed A variable is offered as a source of variation unrelated to unmeasured confounders, and its relevance, its exclusion from the outcome equation and its fit with the question all have to be examined.
A Referee Doubts the Adjustment A reviewer asks whether regression with controls is enough, and the analysis has to show what changes under a weighting estimator, a different comparison group or a sensitivity assessment.
DESIGN AND ESTIMATOR
Treatment Effect Designs in Stata
A command only computes an estimate, and the harder decision is which design the data and the question can sustain.
Defining the Estimand
The average treatment effect, the effect on the treated and a local effect are different quantities, so the target is fixed from the research question before any estimator is chosen.
Adjustment and Weighting
Regression adjustment, inverse-probability weights and doubly robust estimators are fitted with teffects and compared, to see how far the result depends on the outcome model or the allocation model.
Matching Estimators
Nearest-neighbor matching estimators are used where a matched comparison group suits the design, as set out for Propensity Score Matching (PSM).
Instrument-Based Estimates
Estimates from ivregress are examined for the strength and plausibility of the instrument, in line with Instrumental Variable Analysis.
Before-and-After Designs
Policy changes are analyzed by comparing groups before and after the change, with pre-change trends examined, as discussed for Difference-in-Differences (DiD).
Assumption Reporting
Overlap, how similar the groups are after weighting, and the sensitivity of the estimate to unmeasured factors are reported beside each result, not left for the reader to infer.
THREATS TO THE ESTIMATE
How an Estimated Effect Loses Its Footing
Every treatment-effect estimate rests on assumptions the data cannot confirm, so the weaknesses that matter are named and probed instead of hiding behind a single number.
Poor Overlap Between Groups Some participants have almost no counterparts in the other group, so weights become extreme and a few records drive the estimate; overlap is plotted and the target population is stated.
Controls That Cause Bias A variable affected by the treatment is added as a control, which can absorb part of the effect or open a path of bias, so covariates are chosen on timing and reasoning, not on fit.
Spillover to the Comparison Group Treated and untreated people share workplaces, schools or towns, so the program can reach the comparison group indirectly, and the estimate then blends direct and spillover effects unless comparison units are drawn from beyond the program's reach.
Parallel Trends Taken on Trust A policy comparison assumes the groups would have moved together without the change, and a flat pre-period is evidence for that, not proof, especially when the change was anticipated.
Standard Errors Ignore Clusters Policy varies by state or clinic, yet observations are treated as independent, which makes intervals too narrow, so standard errors follow the level at which treatment varies, with small-cluster corrections when the treated units are few.
An Effect Without Its Population An estimate is called the effect of the program, though it applies only to the treated or to a subset whose behavior the instrument changed, so the wording states whom the number describes.
DELIVERABLES
Estimates, Diagnostics and Assumption Notes
The package pairs every estimate with the reasoning and the checks behind it, so that a referee can see what was assumed and what was tested.
Estimand Statement A short statement of the quantity estimated, the population it describes and the assumptions needed for the estimate to carry a causal reading, prepared for the methods section.
Estimator Do-File A do-file that runs each estimator on the analysis file, from the models for allocation or outcome to the final teffects or ivregress call, ready to be rerun.
Overlap Diagnostics Plots and summaries of the estimated probabilities in each group, with the weights or matches that result and any trimming rule that was applied.
Estimator Comparison Table The effect from adjustment, weighting, matching or an instrument set side by side with confidence intervals, so the dependence on method is visible.
Sensitivity to Unmeasured Factors An examination of how far an unmeasured factor would have to act to change the conclusion, or how the estimate moves under alternative comparison groups.
Cautious Wording Note Statistical wording that reports each estimate as an effect for a named population under stated assumptions, matched to the output, for you to check and incorporate.
ILLUSTRATIVE EXAMPLE
An Estimate That Stopped at the Edge of the Data
An illustration of how a program evaluation with no randomization is estimated, with the limits of the comparison reported.
ILLUSTRATIVE EXAMPLE · ADULT EDUCATION POLICY RESEARCH · UNITED STATES
A policy analyst at a research institute wanted to estimate whether enrolling in a workplace literacy program raised later job stability among adult workers, using administrative records and no randomization.
Enrollment depended on prior work history and on local program supply, so allocation was modeled from variables measured before the program, and regression adjustment was kept as a comparison. The complication was overlap: workers with the weakest histories almost never enrolled in some regions, so a few records carried extreme weights. Overlap was plotted, the extreme records were trimmed under a rule fixed in advance, and the estimand was restated for the population that could be compared. The estimate was reported with the share trimmed, a comparison across estimators and wording about an effect conditional on the measured factors.
Outcome: The analyst received the average treatment effect with its 95% confidence interval for each estimator, the overlap plots, the trimming rule, the do-file and a note on the assumptions.
This example does not describe a real client or project. We never share details of actual client work, in line with the confidentiality we promise every researcher. This illustrative example is included to show the type, scope, and standard of work involved in this kind of analysis.
SOFTWARE
Effect Estimation in Stata and Three Alternatives
The estimator matters more than the program, and the same design can be fitted elsewhere when a team's pipeline or a replication request points to another tool.
HOW WE ASSESS YOUR PROJECT
We Review Your Actual Research
Your assessment is based on a review of your actual research materials and analytical requirements, not a generic project category.
You Submit Your Project Details
Share whatever is available at the current stage, such as the research aim, protocol or proposal, dataset, variable list, manuscript or statistical requirements.
We Review Your Research and Define the Scope
Your statistically trained Research Coordinator reviews the materials to understand the study design, research questions, data structure, analytical requirements and expected deliverables. If clarification is needed, additional documents or information can be requested before the assessment is finalized.
Your Analysis Is Completed and Delivered
Once the scope and fixed project fee are agreed, the in-house statistician completes the work and the Research Coordinator remains the communication point through delivery, with the analysis, deliverables and timeline as agreed.
RESEARCH RESPONSIBILITY
Who Stands Behind a Causal Claim
Whether the assumptions behind an estimate are credible in a given setting is a substantive judgment, which the researcher makes and defends in print.
Agreed Statistical Support OnlineSPSS provides the agreed methodological, analytical and statistical reporting support, defined in a scope that is confirmed before any work begins.
No Intellectual Contribution OnlineSPSS does not claim any intellectual contribution to your article, evaluation report or thesis. Authorship remains the researcher's.
Your Research Decisions The program being evaluated, the credibility of the identifying assumptions, the discussion and the conclusions remain yours, with overall responsibility.
Confidential Handling Research data and project details are handled confidentially, project information is not published, and an NDA is available where required.
FREQUENTLY ASKED QUESTIONS
Questions About Treatment Effect Estimation in Stata
A weighting estimator is not automatically safer, not every command takes several treatment levels, and the default estimand may not be the one your question needs.
What is included when OnlineSPSS estimates effects in Stata?
Advanced Stata analysis at OnlineSPSS means estimating treatment or policy effects with an estimator suited to the design, together with the diagnostics and the statement of assumptions that make the estimate interpretable. Adjustment, weighting, matching, instruments and policy-change designs can all be involved. The estimator comes from your question and data, and the wording describes the effect as far as the design allows.
Is doubly robust estimation always safer than regression adjustment?
Not always. A doubly robust estimator stays consistent if either the outcome model or the allocation model is correct, which is a real protection, but it offers none when both are wrong, and poor overlap inflates its variance through extreme weights. It is best run beside a simpler estimator, with the two compared and any large gap investigated.
Can the Stata estimators compare more than two treatment levels?
Some can. Regression adjustment, inverse-probability weighting and the doubly robust estimators in teffects accept a treatment with several levels, such as no program, a short program and a long one, and compare each level with a chosen reference. The matching estimators accept only two groups, so a comparison of several levels is weighted instead or split into pairs, and overlap is checked for every level.
Does Stata report the effect for everyone or only for the treated?
By default the teffects commands estimate the average treatment effect for the whole sample, and the atet option gives the average effect among the treated. The two answer different questions and can differ, so the choice follows the population your question is about. The matching estimators take the same atet option, and the population behind each figure is named in the report.
How do you protect my research data and confidentiality?
Your data and project details are handled confidentially, and project information is not published. Researcher names and institutions are not used publicly without permission. An NDA is available where required and can be arranged before sensitive materials are shared. Read more about confidentiality and data security.
What happens when I request an assessment, and how long does it take?
You share your research question, design and relevant materials, such as the dataset, a description of the program or policy and how participants came to receive it, and any do-files, and a statistically trained Research Coordinator reviews the request and responds the same day. The analysis, deliverables, timeline and one fixed project fee are then confirmed before any work begins, so the timeline follows from the agreed scope.
Still have a question? Find more answers in our full FAQ.
NEXT STEP
Put Your Effect Estimate
on Firmer Ground
How participants came to receive the program, and which outcomes were measured afterward, set the limits of what any estimator can say about its effect.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014