Frequently asked questions

The results of your SudoTrial can help you sell your idea and build your reputation. Amidst a crowded market, customers and investors have increasingly high standards for health and wellness products. We provide methodologically rigorous evidence to support your claims and prove your value.

SudoTrials is for anyone who wants to prove their product or innovation works. SudoTrials are for anyone with data they want to put to work. Don't get stuck in the hassle of dealing with data science or LLM nonsense. Our system quickly and intelligently lays out what to get to answer your questions or what you can reputably do with what you have.

Institutional plans allow organizations to hold one subscription on behalf of multiple users. They are managed by one designated user who can assign three seats to themself and others. Each user has their own personal workspace and usage limit.

Startups and other types of businesses can use the SudoTrials platform to produce evidence to share with potential investors and customers. Showing a product works through reproducible scientific analyses builds trust and supports sales to discerning early users.

SudoTrials are typically retrospective observational trials, that means we use data that was collected previously without a traditional comparison like a drug versus a placebo. Many clinical trials involve recruiting large numbers of participants and monitoring them over time. SudoTrial works with the numbers you already have and the results are ready in minutes.

Approval from a Research Ethics Board or REB is a requirement for any clinical trial involving human participants. Only retrospective studies conducted using deidentified data are exempt. All data that enters the SudoTrial platform is automatically deidentified and anonymized. SudoTrials exclusively use anonymous data and do not require ethics approval.

We automatically deidentify every dataset that is uploaded to our platform. We do not share or sell any data provided by our users.

Our system automatically recognizes identifying information and removes it. Your data is completely anonymized, meaning that it cannot be reidentified. SudoTrials never saves identifying information, and your files remain unchanged in your system.

Example:

User id: John Smith
Email: johnsmith@example.com
Phone number: 555-123-1234

Would become something like:

User id: user-001
Email: user-001-email
Phone number: telephone-number-001

We don't save or use any personal information or personal health information, keeping your data safe and protecting the privacy of your users.

Our standard output template includes a comprehensive methods section, your results, and suitable visualizations. From there, you can share your SudoTrial as a whitepaper and adapt it into social media posts. Your results can be incorporated into sales materials or form the basis of an academic conference presentation or paper.

SudoTrials considers your research goals and the type and amount of data you have. Our suggestions are inspired by previously published studies to ensure they are interesting, novel, and impactful. We are not confined to a particular form of research or type of analysis, we handle qualitative interview studies just as well as we handle building an XGBoost predictive model.

Your priorities may not include bringing in a full-time Data Scientist right now. Some Data Scientists choose to use SudoTrials to generate new ideas or conduct analyses outside of their area of expertise.

There are a number of reasons why you should use caution if you ask ChatGPT or any other generative AI systems to analyze data. These tools may not be equipped to protect your important and sensitive data. Without a detailed, technical prompt, they may default to the simplest or most common methods. The results may change over multiple iterations and depending on the model and version you use. Output from these tools typically requires review from an expert to ensure it is correct, accurate, and free from hallucinations.

SudoTrials was built by experts in experimental design. We use deterministic methods to run all analyses to ensure the results are reliable and reproducible. Our system prioritizes accuracy and clear interpretation.

SudoTrials only suggests analyses that are suitable for your data. It is not always possible to perform a requested analysis with the data provided, but we'll do our best to recommend alternatives and find the best fit.

Rigor refers to the quality of strict precision or exactness. It is generally the goal when designing a scientific experiment and it is a concept we always keep in mind.

SudoTrials exclusively use deidentified data with no possibility of linking to participants. This protects participant privacy. Data is collected by client companies, according to each company's policies.

SudoTrials is compliant with regional privacy laws regarding our users' personal information. SudoTrials does not use or store personal health information collected by clients and does not require a Business Associate Agreement under HIPAA.

We use large language models, LLMs, to help us interpret your input and generate templates. We do not use LLMs to run our analyses, instead we use deterministic methods that produce reproducible results.

Sure, here you go:

Statistical & quantitative (14)

• Hypothesis testing — compare groups or before/after for a difference bigger than chance
• Regression analysis — model how an outcome moves with predictors, holding others fixed
• Bayesian analysis — express results as probabilities given data + prior belief
• Causal inference — estimate an exposure's effect from observational data
• Cluster analysis — group subjects by similarity across measures
• Factor analysis — reduce many correlated items to a few underlying factors
• Latent class analysis — discover hidden subgroups from categorical patterns
• Longitudinal analysis — track how an outcome changes across repeated measurements
• Mediation & moderation analysis — test if an effect runs through/changes with a third variable
• Meta-analysis — combine effect estimates from several studies into one
• Multilevel/hierarchical modeling — model nested data without overcounting
• Structural equation modeling — test a network of relationships at once
• Survival analysis — model time-until-event, handling those who never had it yet
• Time-series analysis — find trend, seasonality, change points over time

AI & Machine Learning (9)

• Classification models — predict a category from features
• Feature importance analysis — rank which inputs most influence an outcome
• Explainable AI analysis — explain what drives a model's predictions
• Fairness & bias assessment — check unequal performance across groups
• Model comparison studies — compare models with honest held-out evaluation
• Robustness evaluation — probe how a model holds up under noise/shift
• Human-AI interaction studies — study how people use/trust an AI system
• Recommendation-system analysis — measure relevance and coverage
• Reinforcement-learning evaluation — evaluate a policy's return vs. a baseline

Natural Language & Text (10)

• Sentiment analysis — score text as positive/negative/neutral
• Topic modeling — discover recurring topics across documents
• Theme extraction — pull salient themes and representative quotes
• Content analysis — code text against a defined scheme and count
• Discourse analysis — interpret how language constructs meaning in context
• Semantic analysis — analyze meaning and similarity between texts
• Linguistic pattern analysis — quantify length/readability/word-category features
• Review mining — extract aspects and opinions from reviews
• Social media analysis — volume, sentiment, themes across posts
• Knowledge extraction — pull structured entities/relations from text

Behavioral Research (9)

• Experiment analysis — analyze a run experiment's outcome between arms
• Engagement analysis — quantify active engagement and what moves it
• Retention analysis — how many users keep coming back over time
• Adoption studies — measure feature/product uptake over time
• Journey analysis — map common user paths through a product
• Clickstream analysis — analyze sequences of page/screen interactions
• User behavior analysis — summarize actions and their relation to outcomes
• Habit-formation analysis — track how a behavior becomes routine
• Decision-making analysis — study how choices are made and what drives them

Economics & outcomes (8)

• Cost-effectiveness analysis — cost per unit of outcome gained (ICER)
• Cost-utility analysis — cost-effectiveness with a QALY-style outcome
• Cost-benefit analysis — costs and benefits in the same units, netted
• Cost-minimization analysis — cheapest option when outcomes are equivalent
• Budget impact analysis — project the total budget effect at scale
• Return on investment — net return relative to cost
• Indirect treatment comparison — compare options never tested head-to-head
• Health resource utilization analysis — summarize visits/admissions/services use

Qualitative research (8)

• Interview analysis — code and summarize themes across transcripts
• Focus group analysis — code themes/dynamics across group discussion
• Thematic coding — systematically tag qualitative data with themes
• Content analysis (also listed above, shared method)
• Case studies — deep structured analysis of one or a few cases
• Ethnographic research — interpret field notes on behavior in context
• Grounded theory development — build theory bottom-up from coded data
• Narrative analysis — analyze structure/meaning of personal stories
• Phenomenological analysis — surface the essence of lived experience

Measurement & Methodology (7)

• Reliability analysis — quantify measurement consistency
• Construct validation — show a measure relates to others as theory predicts
• Method comparison studies — compare two measurement methods for agreement
• Scale development — build and refine a multi-item scale
• Survey design validation — check a survey measures what it should
• Benchmark creation — define a reproducible reference task/metric
• Reproducibility studies — test whether a result reproduces

Experimental Research (6)

• A/B testing — randomly split users, compare an outcome
• Field experiments — run a randomized experiment in a real-world setting
• Intervention studies — introduce an intervention, measure effect over time
• Quasi-experimental studies — estimate effects without randomization

Experimental & observational designs (13)

• Cohort study — follow exposed/unexposed groups forward
• Case-control study — compare those with/without an outcome, looking back
• Cross-sectional study — measure exposure + outcome at one point in time
• Crossover design — each subject receives every condition, as their own control
• Cluster-randomized design — randomize whole groups, not individuals
• Stepped-wedge cluster design — clusters cross to intervention at staggered times
• Dose-response study — test if more exposure yields more/less effect
• Regression discontinuity — exploit a cutoff to estimate a local causal effect
• Non-inferiority/equivalence framing — test an option is no worse within a margin
• Adaptive/group-sequential design — pre-planned interim looks, can stop early
• Diary study — participants log experiences repeatedly in their own context
• Ecological momentary assessment — in-the-moment prompts, many times a day

Looking for something else? Check back soon, this list and our platform's capabilities are always growing.