An AI statistics tutor that draws the curve

It works each idea on a whiteboard step by step, then asks you to interpret the result in your own words, using your own course slides.

Free plan with 50 credits a month. Add your materials once you're in.

  • 3,300+ learners
  • 15,000+ tutoring sessions
  • Cancel anytime
  • Nothing to install

The same lesson, written out

This is a typical exam question because the arithmetic takes two lines and most of the marks are for the sentences around it. On Sitor the tutor builds the test in steps on the whiteboard and asks you to predict each one before it is written.

The question

A coffee machine is set to pour 250 mL. A random sample of 16 cups has a mean of 246 mL and a standard deviation of 8 mL, and a dot plot of the 16 volumes shows no strong skew or outliers. At α = 0.05, is there evidence that the mean pour differs from 250 mL? Explain in words what the p-value means.

Your turnCheck yourself: suppose the same mean (246 mL) and standard deviation (8 mL) came from 25 cups instead of 16. Does the conclusion change, and why?Show the answer

Yes. The standard error becomes 8/√25 = 1.6 mL, so t = (246 − 250) / 1.6 = −2.50 with 24 degrees of freedom, and the two-sided p-value is about 0.020. That is below 0.05, so we reject H₀: there is convincing evidence that the mean pour differs from 250 mL. The 4 mL difference is the same; the larger sample made the standard error smaller, so the same difference is stronger evidence.

  1. Say what is being tested

    Hypotheses are about the parameter μ, the mean of all cups the machine pours, not about the sample mean. H₀: μ = 250 mL. Hₐ: μ ≠ 250 mL. It is two-sided because the question asks whether the mean "differs", not whether it is lower.

  2. Choose the test and check conditions

    One sample, one quantitative variable, and σ is unknown, so this is a one-sample t-test with 16 − 1 = 15 degrees of freedom. The sample is random, and with only 16 cups the data need to look roughly normal, which the dot plot supports.

  3. Standard error and test statistic

    The standard error of the mean is s/√n = 8/√16 = 2 mL. The test statistic is t = (x̄ − 250) / 2 = (246 − 250) / 2 = −2.00. The sample mean is 2 standard errors below the claimed value.

  4. Find the p-value

    On a t curve with 15 degrees of freedom, the area below −2.00 plus the area above +2.00 is about 0.064 (0.0639 to four decimal places). Both tails are shaded because Hₐ is two-sided.

  5. Decide and conclude in context

    Since 0.064 is greater than α = 0.05, we fail to reject H₀. The data do not give convincing evidence that the mean pour differs from 250 mL. This does not show that the machine is set correctly: if its true mean is not 250 mL, this decision is a Type II error, and a sample of 16 cups has limited power to detect a 4 mL difference.

  6. Say what the p-value means

    If the machine really did average 250 mL, about 6.4% of random samples of 16 cups would give a t statistic at least as far from zero as −2.00. It is not a 6.4% chance that H₀ is true. As a cross-check, the 95% confidence interval is 246 ± 2.131 × 2, or about 241.7 mL to 250.3 mL, and it contains 250.

What makes intro statistics hard, and how Sitor teaches it

Most points in a first statistics course are lost on interpretation, not calculation. A calculator gives the number; the exam asks what the number means. These are the six ideas students most often get wrong in words.

TopicWhere students lose pointsHow Sitor teaches it
Conditional probability and base ratesStudents treat P(A | B) and P(B | A) as the same number, so "the test is 90% accurate" turns into "a positive result means a 90% chance of having the disease".The board fills in a table of counts for 1,000 people instead of a formula. With 2% of people sick and a test that is right 90% of the time for both sick and healthy people, 18 of the 116 positives are real, about 16%. You are then asked which row or column total is the denominator for each question.
Sampling distributions and the Central Limit TheoremThe distribution of the data gets confused with the distribution of the sample mean. Students believe a large sample makes the data themselves normal, and use σ where σ/√n belongs.Three sketches are stacked on the board: the population, one sample, and the distribution of x̄ over many samples with its spread labeled σ/√n. You are asked what happens to each sketch when n goes from 25 to 100.
Confidence intervals"95% confident" is read as "95% of the data are in the interval" or as "95% of sample means fall in this interval". The level describes how often the method captures the parameter, not the data and not other samples.The tutor draws 20 intervals from 20 samples around a fixed μ, with about one missing it, to show what the 95% refers to. Then you write the interpretation of a new interval and it is checked for three things: the parameter, the context and the confidence level.
P-values and conclusionsThe p-value is read as "the probability that the null hypothesis is true", and "fail to reject H₀" is written as "accept H₀" or "this proves there is no difference".A curve is drawn centered on the null value and the area beyond the observed statistic is shaded. That area is the p-value, computed assuming H₀ is true. You then write the full conclusion: the comparison with α, the decision, and what it says about the claim in context.
Type I and Type II errors, and powerThe definitions are memorized as "false positive" and "false negative", but students cannot describe either error in the context of the study, or explain why making α smaller has a cost.A two-by-two grid of truth against decision is filled in using the words of the actual problem. Two overlapping curves then show that moving the cutoff to shrink α makes a Type II error more likely, and that a larger sample narrows both curves and raises power.
Correlation, regression and causationThe slope is described as a guaranteed change for one individual, r is confused with r², predictions are made far outside the data, and a strong correlation is taken as proof of cause.On a scatterplot with its fitted line you write the slope sentence yourself ("for each additional hour, the predicted score is higher by b points, on average"), read r² as the share of variation in y explained by the line, then name a possible lurking variable and the study design that would allow a causal claim.

A three-week study plan for a statistics exam

A plan for a statistics midterm, final or the AP Statistics exam that covers probability through inference. Tell Sitor your exam date and it builds this from your own syllabus; the shape stays the same.

  1. 3 weeks out

    Find the gaps

    • Upload the lecture slides for every unit on the exam, plus the formula sheet and tables your course allows.
    • Take the diagnostic and answer "not sure" honestly, especially on the questions that ask for an interpretation.
    • Repair the foundations first if they are weak: z-scores, the normal curve and conditional probability. Inference is built on them.
  2. 2 weeks out

    Build the chain of inference

    • Work in order on the whiteboard: sampling distributions, then confidence intervals, then hypothesis tests. Each one uses the one before.
    • For every procedure, write four things by hand: the hypotheses or parameter, the conditions, the calculation and the conclusion in context.
    • Do not move on from a topic until the tutor marks it mastered.
  3. 1 week out

    Practice choosing the test

    • Ask for mixed questions written from your slides, where nothing tells you whether it is a t-test, a proportion test, chi-square or regression.
    • Practice reading results from the tool your exam uses: calculator, R or SPSS output, or printed tables.
    • Work any past papers your course has published, timed, and mark the interpretation sentences as strictly as the calculations.
  4. The last two days

    Review, do not cram

    • Do the spaced reviews waiting on your Today page.
    • Write from memory one model sentence each for a confidence interval, a p-value, a test conclusion and a regression slope.
    • Sleep. A tired reader misses "paired" or "at least" in the question, and that costs more than one more hour of practice gains.

Common mistakes when studying statistics

Stopping at the number
Getting t = −2.00 feels like finishing the problem. The conclusion in context is often worth as many marks as the calculation, so write the full sentence every time you practice.
Practicing one chapter at a time
In the chapter on t-tests every question is a t-test, so you never practice choosing. The exam mixes them. Spend the last week on mixed sets where picking the procedure is the first step.
Memorizing formulas that are on the formula sheet
Many courses, and the AP exam, hand you the formulas. What they do not hand you is which one applies, what each symbol stands for, and which conditions must hold. Study those.
Letting the calculator do the thinking
A test menu will return a p-value for the wrong test as readily as for the right one. Sketch the curve and shade the tail first, so you can tell when an answer is one-sided, two-sided or simply implausible.

Start free. Upgrade when it earns its place.

  • Free

    $0

    Monthly credits
    50/mo
    Learning topics
    Up to 10

    Enough to work through a few topics and see if it suits how you study.

  • Pro

    $19.90/mo

    First month $9.90, then $19.90/mo

    Monthly credits
    1,000/mo
    Learning topics
    Up to 500

    For exam season: the most credits plus deep research.

Start free

Questions about studying statistics with an AI tutor

Can an AI tutor really teach college statistics?

For the concepts and procedures of an intro course, yes: they are well-established, and an AI tutor can explain, draw and question you on them at any hour. Like any AI, Sitor can make an arithmetic slip, so check final numbers against your calculator, software or tables. It does not replace your professor and it is not an official question bank; keep using the past papers your course provides.

Does it work with my course, such as STAT 101, STAT 200, statistics for psychology or AP Statistics?

Yes. Upload the slides, notes or textbook chapters your course uses and Sitor builds the study plan and the practice questions from those pages, so the notation, the scope and the wording of conclusions match what you will be examined on. Without uploads it teaches from a standard introductory statistics syllabus.

How is this different from asking ChatGPT to solve a statistics problem?

A chatbot answers the question you ask, and a solved problem is not the same as being able to solve the next one. Sitor starts with a diagnostic, plans the whole syllabus, teaches each topic step by step on a whiteboard, and asks you to write the interpretation yourself. It keeps track of what you have mastered and brings topics back for review before you forget them.

How do I know which statistical test to use?

Ask three questions. Is the variable categorical or quantitative? How many groups or variables are there? Are the samples independent or paired? One mean is a one-sample t-test, two independent groups a two-sample t-test, the same subjects measured twice a paired t-test, proportions a z-test, two categorical variables a chi-square test, and two quantitative variables regression. Sitor drills this with mixed questions where you choose the test before any calculation.

Can it help with R, SPSS or calculator output?

Yes, with reading it. Paste the output into the chat or upload a photo of it, and the tutor explains what each line means (the test statistic, degrees of freedom, p-value and confidence interval) and how to write the result up. Sitor is a tutor, not statistical software, so run the analysis in the tool your course uses and treat that output as the number of record.

How much does it cost?

There is a free plan with 50 credits a month. A turn in the text chat starts at 1 credit and a turn on the whiteboard uses 10, so the free plan covers a short whiteboard lesson or a longer chat session. Starter is $9.90 a month and Pro is $19.90 a month, with a discount on the first month. You can cancel at any time.

Which part of statistics are you studying?

Free plan with 50 credits a month. Add your materials once you're in.