Natural Language Statistics Suite

Statistics,
in conversation.

Ask your local AI agent to analyse your data. NLSS runs the statistics in R and saves results you can inspect.

Open source. Free to use under the Apache 2.0 license.

Install NLSS

Choose your agent and follow the setup guide.

Start tutorial

Explore a learning study with guided prompts.

ResearcherCompare test scores after active practice and rereading.

NLSSIn this simulated study, practice scores averaged 5.65 points higher. The 95% confidence interval for the difference is 0.86–10.44 points. Shall we look at how much the scores overlap?

Learning strategies

Analysis report · calculated with R

Descriptive statistics for synthetic learning scores
ConditionnMeanSD
Practice4067.4710.19
Rereading4061.8311.30
+5.65score points
95% CI [0.86, 10.44]

Synthetic data · 80 learners. A result to interpret, not evidence about real learning strategies.

About this example

The conversation is illustrative. The numbers are verified results from the tutorial’s simulated learning study. View the tutorial results.

How NLSS works

From your question
to an answer you can check.

Talk through your study. Your agent connects the conversation to NLSS’s fixed R routines, then brings the results back to you.

  1. 01 · You

    Ask your question

    Describe what you want to learn and share the study context and data in your local folder.

    Your research question
  2. 02 · Your agent

    Make an analysis plan

    The agent clarifies the question, discusses the method with you and calls the matching NLSS procedure.

    A method and its settings
  3. 03 · NLSS + R

    Calculate and record

    Fixed R scripts do the calculation. NLSS scripts save the inputs, settings and results, and build the analysis protocol. This happens locally in your folder.

    See an analysis protocol
  4. 04 · Your agent

    Explain the findings

    The agent reads the saved results, answers your question and writes a report when you ask.

    See an example report

You stay responsible for the method and interpretation. The saved analysis gives you something concrete to check, discuss and return to.

Conversation, with a dependable calculation behind it.

NLSS brings reusable R procedures and an automatic analysis record into the same conversation. You can focus on the study, with less code and record-keeping to assemble yourself.

How four analysis workflows organize calculation and records
Your starting pointHow you work
Statistical softwareChoose procedures through menus or syntax. Keep the syntax and output so you can revisit the analysis.
Statistics means navigating a complex user interface.
Write R yourselfBuild and maintain your scripts, and organize the data, results and reports around them.
Statistics means editing code and managing files.
AI-generated R codeGenerate code for your question, then review, run and preserve the version you used.
Statistics means engineering prompts.
Work with NLSSDiscuss your question. The agent calls fixed R routines, with inputs, settings, results and a protocol saved together.
No complex interface to navigate. No code to write. No elaborate prompts. Just conversation.
What makes the calculation repeatable?

NLSS separates the conversation from the numerical calculation. Its R routines use defined inputs and settings. Repeating a saved analysis requires the same data, options and compatible software; methods involving randomness also need a fixed seed. The agent’s wording and choices in a new conversation can vary.

Statistical software and well-maintained R projects can also support reproducible work. The comparison describes workflows: asking an AI assistant for new R code is different from using a saved analysis routine. A recorded calculation still needs a suitable method and careful interpretation.

Read more: NLSS protocols and reports · Repeating saved NLSS analyses

Try it yourself

Tutorial: your first study with NLSS.

Would you learn more by practising or by reading again? Explore a simulated experiment with 80 learners, compare the groups and write a short report.

Download the study files, then work through five prompts in your own agent. Each step includes a question to consider and a reference answer to compare.

Start tutorial

Five steps, from data to report

  1. Meet the data

    Understand the study and explore the scores.

  2. Discuss the analysis

    Decide how to compare the groups.

  3. Estimate the difference

    Read the effect alongside its uncertainty.

  4. Challenge the conclusion

    Ask what the result really supports.

  5. Write and inspect

    Connect your report to the saved evidence.