Skip to content

CodeSplash Analytics v0.9.0

Choose a tag to compare

@codesplash-ai codesplash-ai released this 12 Nov 00:22
· 1 commit to main since this release

Learning, built into the instrument

CodeSplash now teaches as you work. Every test in the statistics guide has a page you can read at your depth, every result card can explain itself, and walkthroughs show you a page by having you do the real task on it.

Learn

  • CodeSplash Academy. A new section where every test in the statistics guide has its own page: the question it answers, its hypotheses and assumptions, the formula, how to read the output, pitfalls, and checked Python and R code. Choosing a method is the hard part. Each page says when the test fits and when it doesn't, and where CodeSplash runs it, one click sets it up in Analyze.
  • Tracks and lessons. Lessons grouped by school and track, each with something to try, checks along the way and a real run at the end. Pick the path that fits your job and the list follows it. Short lessons with practice built in stick better than reading a manual once.
  • Key and Atlas. The statistics guide as a route map (Key) and the data science and ML guide as an atlas. Trace a route with your own data and the map shows what fits. You can start from the question you have, not from the name of a test.
  • Guided walkthroughs. A walkthrough points at the real controls and moves on when you do the step. Hints, Show me and Do it for me are there when you're stuck. Run it in this project, in the practice project, or on your own columns. You learn a page by using it, and what you make along the way is yours to keep.
  • Simulation Lab. Stress-test a method on data where you know the truth: how often it rejects, how far off its estimates are, and how often its intervals cover the true value. Seeing a test fail on purpose is the quickest way to trust it when it passes.
  • Pitfalls, Notation and I need to…. The common misconceptions with what's true instead, what each symbol means and how to say it, and a way in that starts from the job you have to get done. The mistakes smart people make are worth meeting before they cost you a conclusion.

On every page

  • Explain mode. Turn it on and every term, number and control that can explain itself is underlined, until you switch it off. The explanation sits where the question comes up, on your own screen.
  • Learn from a result. The graduation cap on a result card holds Teach me (numbered marks in reading order), Understand this result, Change one thing, Check my interpretation, and Predict before each run. A result card says a lot at once. These show you what to read first and what each number means.
  • Worked examples and sample data. Each method in Analyze can load its own worked examples, set up on licensed sample data marked “(sample)”. Empty pages offer a sample dataset too, so you can try a page before you bring data. A worked example shows the setup that a method expects, on data where the answer is known.

Studios

  • Meta-analysis studio. Pool effects across studies with common- or random-effects models, explore why studies differ and whether small studies tell a different story, and combine p-values. Evidence from several studies deserves more than a vote count.
  • Spatial studio. Attach geometry to your data, define who neighbours whom, test for clusters and hot spots, analyze point patterns and fit spatial regressions. Places near each other tend to be alike, and ordinary tests assume they aren't.
  • Series lab. Look at a time series, test stationarity, dependence, trends and breaks, then hand it to forecasting. A forecast is only as good as your understanding of the series behind it.

Getting started

  • A Home that knows what's next. Home suggests one next step from what your project has, picks up where you left off, and shows today's practice, your skill map, recent work, and whether Python, R and AI are ready. Opening the app should tell you where to go, not ask you.
  • Set up in six steps. A short setup asks what you're here for, checks Python and R, offers a local AI model, and ends with your first result: a Welch t test on a small experiment, read with numbered marks. A real result in the first minute beats a tour of buttons. Help ▸ Take the Tour Again brings it back.
  • Sample projects. Help ▸ Sample Projects opens a small project built around one licensed dataset, with a page to start on and a first thing to try. Practicing on data that isn't yours means nothing you try can break your work.
  • Report a Problem. Help ▸ Report a Problem puts together what you were doing, what happened, and the facts about this computer and this build that a bug report needs. Nothing is sent from the app. Copy the report, save it as a file, or open it as a draft email or GitHub issue, and you choose whether to send it.

Download CodeSplash Analytics at https://analytics.codesplash.ai/download. To check a download, compare it with SHA256SUMS below.