Binomial Test

download Binomial Test

of 5

Transcript of Binomial Test

  • 8/13/2019 Binomial Test

    1/5

    SM439: Nonparametric Statistics Prof. Jager

    Binomial Test

    Example: (Rossman and Chance, 2008) The ancient Greeksextensively used the golden ratio in art and literature

    they believed that a width-to-length ratio of 1+

    52

    = 0.618

    was aesthetically pleasing. Some have conjectured that

    American Indians also used this ratio (Hand et al., 1993).

    The following data are width-to-length ratios for a random

    sample of 20 beaded rectangles used by the Shoshoni

    Indians to decorate their leather goods.

    0.693 0.662 0.690 0.606 0.570

    0.749 0.672 0.628 0.609 0.844

    0.654 0.615 0.668 0.601 0.576

    0.670 0.606 0.611 0.553 0.933

    Use this data to determine whether there is evidence that

    the Shoshoni Indians also employed the golden ratio in

    their art.

    What are we trying to do here?

    Old Way: Parametric methods

    New Way: Nonparametric methods BINOMIAL TEST!

    USNA 26 Aug 2009 1

  • 8/13/2019 Binomial Test

    2/5

    SM439: Nonparametric Statistics Prof. Jager

    Old Way: One-sample t-test

    Let =

    We want to test the following hypotheses:

    Our test statistic is:

    In our sample:

    If H0 is true, then:

    So ourp

    -value is:

    Conclusion:

    Assumptions:

    USNA 26 Aug 2009 2

  • 8/13/2019 Binomial Test

    3/5

    SM439: Nonparametric Statistics Prof. Jager

    New Way: Binomial Test

    Test uses the median instead of the mean.

    For a continuous distribution, the median (.5) satisfies:

    P r(X > .5) =P r(X < .5) =.5

    For the golden rule example, our hypotheses are then:

    Our tests statistic is: B =

    IfH0 is true, then the distribution ofB is:

    Our p-value is:

    Conclusion:

    USNA 26 Aug 2009 3

  • 8/13/2019 Binomial Test

    4/5

    SM439: Nonparametric Statistics Prof. Jager

    Normal Approximation for Binomial Test

    When the sample is large enough (n 30), we can use the

    normal approximation to the binomial distribution to

    perform this test:

    Note: Since we have computers, it is just as easy to

    calculate the p-value exactly, rather than approximately!

    For the approximation, our test statistic is:

    For our golden rule example:

    USNA 26 Aug 2009 4

  • 8/13/2019 Binomial Test

    5/5

    SM439: Nonparametric Statistics Prof. Jager

    Using R

    The sign.test() function from the BSDA library will

    perform a binomial test for us:

    > library(BSDA)

    > goldenratio = read.table(file.choose(), header=T)

    > names(goldenratio)

    [1] "ratios"

    > attach(goldenratio)

    > sign.test(ratios, md = .618, alternative="two.sided")

    $rval

    One-sample Sign-Test

    data: ratios

    s = 11, p-value = 0.8238

    alternative hypothesis: true median is not equal to 0.618

    95 percent confidence interval:

    0.6063494 0.6717671

    sample estimates:

    median of x

    0.641

    $Confidence.Intervals

    Conf.Level L.E.pt U.E.pt

    Lower Achieved CI 0.8847 0.6090 0.6700

    Interpolated CI 0.9500 0.6063 0.6718

    Upper Achieved CI 0.9586 0.6060 0.6720

    USNA 26 Aug 2009 5