FrançaisGet started

Statistical analysis libraries

Doing the path's calculations in Python: scipy.stats (quantiles, t.interval, t-tests, chi-squared), statsmodels for a proportion's interval, pandas for the contingency table; the R equivalents are named.

  • 22 cards
  • 5 topics
  • Free
Study this deck

The outline, topic by topic

Each topic is weighted the way the exam weighs it.

  1. Concepts
  2. Questions
  3. Sort and link
  4. Scenarios
  5. Understand

A few cards from the deck

  • scipy.stats: what is it?

    Show the answer

    A Python scientific computing module that provides probability distributions (norm, t, chi2) and statistical tests (ttest_1samp, ttest_ind, ttest_rel, chisquare, chi2_contingency).

    Each distribution offers ppf (quantile), cdf (cumulative probability) and sf (tail). Each test returns the statistic and the p-value.

  • What does this program print?

    ```python
    from scipy import stats
    print(round(stats.norm.ppf(0.975), 2))
    `

    • 1.96
    • 1.64
    • 0.98
    • 2.58
    Show the answer

    1.96

    ppf(0.975) leaves 2.5% of the normal distribution on the right: it is the quantile of the two-sided 95% interval. 1.64 is ppf(0.95) and 2.58 is ppf(0.995).

  • In R, the function … runs the t-tests, … the chi-squared test and … tests a proportion.

    Show the answer

    In R, the function t.test runs the t-tests, chisq.test the chi-squared test and prop.test tests a proportion.

    t.test(x, y, paired = TRUE) corresponds to scipy's ttest_rel. R gathers under one name what scipy splits into several functions.

  • statsmodels: what is it?

    Show the answer

    A Python statistical analysis library, offering among other things the confidence interval of a proportion with proportion_confint.

    proportion_confint(successes, total, alpha=0.05, method='normal') returns both bounds. The method option picks the formula, 'normal' being the one in the course.

  • The same 30 pages are measured before and after an optimisation, and the script runs stats.ttest_ind(before, after). Which fix is right?

    • stats.ttest_1samp(before, after)
    • stats.chisquare(before, after)
    • stats.ttest_rel(before, after), paired series
    • stats.ttest_ind(before, after, equal_var=False)
    Show the answer

    stats.ttest_rel(before, after), paired series

    ttest_rel tests page-by-page differences. equal_var=False fixes another problem (unequal variances between two distinct groups). ttest_1samp expects a value, not a second series.

Study this deck