How to infer difference of population proportion between two groups when proportion is small?





.everyoneloves__top-leaderboard:empty,.everyoneloves__mid-leaderboard:empty,.everyoneloves__bot-mid-leaderboard:empty{ margin-bottom:0;
}







2












$begingroup$


I have a dataset where the issue is of this form.



There are two groups, Group A (N=5000) and Group B (N=1000). Let's say 5 people in Group A develop a certain disease, and only 2 in group B do.



Then the proportion for A is 5/5000 -> 0.001 and for B it is 2/1000 -> 0.002.



How can I test if the proportion between these two groups is statistically significant?



The tests I found online rely on the Central Limit Theorem, such that np>=10 and n(1-p) >= 10, which does not hold for my dataset. Are there any other approaches?










share|cite|improve this question







New contributor




max is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.







$endgroup$












  • $begingroup$
    Use Fisher Exact Test per discussion in Answer.
    $endgroup$
    – BruceET
    Apr 22 at 2:38


















2












$begingroup$


I have a dataset where the issue is of this form.



There are two groups, Group A (N=5000) and Group B (N=1000). Let's say 5 people in Group A develop a certain disease, and only 2 in group B do.



Then the proportion for A is 5/5000 -> 0.001 and for B it is 2/1000 -> 0.002.



How can I test if the proportion between these two groups is statistically significant?



The tests I found online rely on the Central Limit Theorem, such that np>=10 and n(1-p) >= 10, which does not hold for my dataset. Are there any other approaches?










share|cite|improve this question







New contributor




max is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.







$endgroup$












  • $begingroup$
    Use Fisher Exact Test per discussion in Answer.
    $endgroup$
    – BruceET
    Apr 22 at 2:38














2












2








2


1



$begingroup$


I have a dataset where the issue is of this form.



There are two groups, Group A (N=5000) and Group B (N=1000). Let's say 5 people in Group A develop a certain disease, and only 2 in group B do.



Then the proportion for A is 5/5000 -> 0.001 and for B it is 2/1000 -> 0.002.



How can I test if the proportion between these two groups is statistically significant?



The tests I found online rely on the Central Limit Theorem, such that np>=10 and n(1-p) >= 10, which does not hold for my dataset. Are there any other approaches?










share|cite|improve this question







New contributor




max is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.







$endgroup$




I have a dataset where the issue is of this form.



There are two groups, Group A (N=5000) and Group B (N=1000). Let's say 5 people in Group A develop a certain disease, and only 2 in group B do.



Then the proportion for A is 5/5000 -> 0.001 and for B it is 2/1000 -> 0.002.



How can I test if the proportion between these two groups is statistically significant?



The tests I found online rely on the Central Limit Theorem, such that np>=10 and n(1-p) >= 10, which does not hold for my dataset. Are there any other approaches?







inference proportion






share|cite|improve this question







New contributor




max is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.











share|cite|improve this question







New contributor




max is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.









share|cite|improve this question




share|cite|improve this question






New contributor




max is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.









asked Apr 21 at 22:52









maxmax

1133




1133




New contributor




max is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.





New contributor





max is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.






max is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.












  • $begingroup$
    Use Fisher Exact Test per discussion in Answer.
    $endgroup$
    – BruceET
    Apr 22 at 2:38


















  • $begingroup$
    Use Fisher Exact Test per discussion in Answer.
    $endgroup$
    – BruceET
    Apr 22 at 2:38
















$begingroup$
Use Fisher Exact Test per discussion in Answer.
$endgroup$
– BruceET
Apr 22 at 2:38




$begingroup$
Use Fisher Exact Test per discussion in Answer.
$endgroup$
– BruceET
Apr 22 at 2:38










1 Answer
1






active

oldest

votes


















2












$begingroup$

The traditional approach would be to do a 2-sample test for a difference in proportions: In Minitab, results of this test are as shown below. The warning about the first P-value from a normal approximation causes doubt,
roughly for the reasons you mention.



However, the result from Fisher's exact test uses an exact hypergeometric
probability. It also shows no significant difference.



Test and CI for Two Proportions 

Sample X N Sample p
1 5 5000 0.001000
2 2 1000 0.002000

Difference = p (1) - p (2)
Estimate for difference: -0.001
95% upper bound for difference: 0.00143738
Test for difference = 0 (vs < 0):
Z = -0.67 P-Value = 0.250

* NOTE * The normal approximation may be
inaccurate for small samples.

Fisher’s exact test: P-Value = 0.330


A direct hypergeometric computation in R can be argued
as follows. Suppose an urn contains $5000$ tokens marked A and $1000$ marked B. Seven tokens are taken
at random without replacement, corresponding to disease.
What is the probability five or fewer of those are marked A?



The answer is



$$sum_{k=0}^5frac{{5000 choose k}{1000 choose 7-k}}{{6000 choose 7}} = 0.3302,$$



which agrees with the P-value from Fisher's exact test.



In R, the computation can be done in terms of a hypergeometric CDF:



phyper(5, 5000, 1000, 7)
[1] 0.330204


Here is a plot of this hypergeometric distribution. The P-value is the sum of the heights of the bars to the left of the vertical dotted line.



enter image description here






share|cite|improve this answer









$endgroup$














    Your Answer








    StackExchange.ready(function() {
    var channelOptions = {
    tags: "".split(" "),
    id: "65"
    };
    initTagRenderer("".split(" "), "".split(" "), channelOptions);

    StackExchange.using("externalEditor", function() {
    // Have to fire editor after snippets, if snippets enabled
    if (StackExchange.settings.snippets.snippetsEnabled) {
    StackExchange.using("snippets", function() {
    createEditor();
    });
    }
    else {
    createEditor();
    }
    });

    function createEditor() {
    StackExchange.prepareEditor({
    heartbeatType: 'answer',
    autoActivateHeartbeat: false,
    convertImagesToLinks: false,
    noModals: true,
    showLowRepImageUploadWarning: true,
    reputationToPostImages: null,
    bindNavPrevention: true,
    postfix: "",
    imageUploader: {
    brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
    contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
    allowUrls: true
    },
    onDemand: true,
    discardSelector: ".discard-answer"
    ,immediatelyShowMarkdownHelp:true
    });


    }
    });






    max is a new contributor. Be nice, and check out our Code of Conduct.










    draft saved

    draft discarded


















    StackExchange.ready(
    function () {
    StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstats.stackexchange.com%2fquestions%2f404331%2fhow-to-infer-difference-of-population-proportion-between-two-groups-when-proport%23new-answer', 'question_page');
    }
    );

    Post as a guest















    Required, but never shown

























    1 Answer
    1






    active

    oldest

    votes








    1 Answer
    1






    active

    oldest

    votes









    active

    oldest

    votes






    active

    oldest

    votes









    2












    $begingroup$

    The traditional approach would be to do a 2-sample test for a difference in proportions: In Minitab, results of this test are as shown below. The warning about the first P-value from a normal approximation causes doubt,
    roughly for the reasons you mention.



    However, the result from Fisher's exact test uses an exact hypergeometric
    probability. It also shows no significant difference.



    Test and CI for Two Proportions 

    Sample X N Sample p
    1 5 5000 0.001000
    2 2 1000 0.002000

    Difference = p (1) - p (2)
    Estimate for difference: -0.001
    95% upper bound for difference: 0.00143738
    Test for difference = 0 (vs < 0):
    Z = -0.67 P-Value = 0.250

    * NOTE * The normal approximation may be
    inaccurate for small samples.

    Fisher’s exact test: P-Value = 0.330


    A direct hypergeometric computation in R can be argued
    as follows. Suppose an urn contains $5000$ tokens marked A and $1000$ marked B. Seven tokens are taken
    at random without replacement, corresponding to disease.
    What is the probability five or fewer of those are marked A?



    The answer is



    $$sum_{k=0}^5frac{{5000 choose k}{1000 choose 7-k}}{{6000 choose 7}} = 0.3302,$$



    which agrees with the P-value from Fisher's exact test.



    In R, the computation can be done in terms of a hypergeometric CDF:



    phyper(5, 5000, 1000, 7)
    [1] 0.330204


    Here is a plot of this hypergeometric distribution. The P-value is the sum of the heights of the bars to the left of the vertical dotted line.



    enter image description here






    share|cite|improve this answer









    $endgroup$


















      2












      $begingroup$

      The traditional approach would be to do a 2-sample test for a difference in proportions: In Minitab, results of this test are as shown below. The warning about the first P-value from a normal approximation causes doubt,
      roughly for the reasons you mention.



      However, the result from Fisher's exact test uses an exact hypergeometric
      probability. It also shows no significant difference.



      Test and CI for Two Proportions 

      Sample X N Sample p
      1 5 5000 0.001000
      2 2 1000 0.002000

      Difference = p (1) - p (2)
      Estimate for difference: -0.001
      95% upper bound for difference: 0.00143738
      Test for difference = 0 (vs < 0):
      Z = -0.67 P-Value = 0.250

      * NOTE * The normal approximation may be
      inaccurate for small samples.

      Fisher’s exact test: P-Value = 0.330


      A direct hypergeometric computation in R can be argued
      as follows. Suppose an urn contains $5000$ tokens marked A and $1000$ marked B. Seven tokens are taken
      at random without replacement, corresponding to disease.
      What is the probability five or fewer of those are marked A?



      The answer is



      $$sum_{k=0}^5frac{{5000 choose k}{1000 choose 7-k}}{{6000 choose 7}} = 0.3302,$$



      which agrees with the P-value from Fisher's exact test.



      In R, the computation can be done in terms of a hypergeometric CDF:



      phyper(5, 5000, 1000, 7)
      [1] 0.330204


      Here is a plot of this hypergeometric distribution. The P-value is the sum of the heights of the bars to the left of the vertical dotted line.



      enter image description here






      share|cite|improve this answer









      $endgroup$
















        2












        2








        2





        $begingroup$

        The traditional approach would be to do a 2-sample test for a difference in proportions: In Minitab, results of this test are as shown below. The warning about the first P-value from a normal approximation causes doubt,
        roughly for the reasons you mention.



        However, the result from Fisher's exact test uses an exact hypergeometric
        probability. It also shows no significant difference.



        Test and CI for Two Proportions 

        Sample X N Sample p
        1 5 5000 0.001000
        2 2 1000 0.002000

        Difference = p (1) - p (2)
        Estimate for difference: -0.001
        95% upper bound for difference: 0.00143738
        Test for difference = 0 (vs < 0):
        Z = -0.67 P-Value = 0.250

        * NOTE * The normal approximation may be
        inaccurate for small samples.

        Fisher’s exact test: P-Value = 0.330


        A direct hypergeometric computation in R can be argued
        as follows. Suppose an urn contains $5000$ tokens marked A and $1000$ marked B. Seven tokens are taken
        at random without replacement, corresponding to disease.
        What is the probability five or fewer of those are marked A?



        The answer is



        $$sum_{k=0}^5frac{{5000 choose k}{1000 choose 7-k}}{{6000 choose 7}} = 0.3302,$$



        which agrees with the P-value from Fisher's exact test.



        In R, the computation can be done in terms of a hypergeometric CDF:



        phyper(5, 5000, 1000, 7)
        [1] 0.330204


        Here is a plot of this hypergeometric distribution. The P-value is the sum of the heights of the bars to the left of the vertical dotted line.



        enter image description here






        share|cite|improve this answer









        $endgroup$



        The traditional approach would be to do a 2-sample test for a difference in proportions: In Minitab, results of this test are as shown below. The warning about the first P-value from a normal approximation causes doubt,
        roughly for the reasons you mention.



        However, the result from Fisher's exact test uses an exact hypergeometric
        probability. It also shows no significant difference.



        Test and CI for Two Proportions 

        Sample X N Sample p
        1 5 5000 0.001000
        2 2 1000 0.002000

        Difference = p (1) - p (2)
        Estimate for difference: -0.001
        95% upper bound for difference: 0.00143738
        Test for difference = 0 (vs < 0):
        Z = -0.67 P-Value = 0.250

        * NOTE * The normal approximation may be
        inaccurate for small samples.

        Fisher’s exact test: P-Value = 0.330


        A direct hypergeometric computation in R can be argued
        as follows. Suppose an urn contains $5000$ tokens marked A and $1000$ marked B. Seven tokens are taken
        at random without replacement, corresponding to disease.
        What is the probability five or fewer of those are marked A?



        The answer is



        $$sum_{k=0}^5frac{{5000 choose k}{1000 choose 7-k}}{{6000 choose 7}} = 0.3302,$$



        which agrees with the P-value from Fisher's exact test.



        In R, the computation can be done in terms of a hypergeometric CDF:



        phyper(5, 5000, 1000, 7)
        [1] 0.330204


        Here is a plot of this hypergeometric distribution. The P-value is the sum of the heights of the bars to the left of the vertical dotted line.



        enter image description here







        share|cite|improve this answer












        share|cite|improve this answer



        share|cite|improve this answer










        answered Apr 22 at 2:07









        BruceETBruceET

        7,2861721




        7,2861721






















            max is a new contributor. Be nice, and check out our Code of Conduct.










            draft saved

            draft discarded


















            max is a new contributor. Be nice, and check out our Code of Conduct.













            max is a new contributor. Be nice, and check out our Code of Conduct.












            max is a new contributor. Be nice, and check out our Code of Conduct.
















            Thanks for contributing an answer to Cross Validated!


            • Please be sure to answer the question. Provide details and share your research!

            But avoid



            • Asking for help, clarification, or responding to other answers.

            • Making statements based on opinion; back them up with references or personal experience.


            Use MathJax to format equations. MathJax reference.


            To learn more, see our tips on writing great answers.




            draft saved


            draft discarded














            StackExchange.ready(
            function () {
            StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstats.stackexchange.com%2fquestions%2f404331%2fhow-to-infer-difference-of-population-proportion-between-two-groups-when-proport%23new-answer', 'question_page');
            }
            );

            Post as a guest















            Required, but never shown





















































            Required, but never shown














            Required, but never shown












            Required, but never shown







            Required, but never shown

































            Required, but never shown














            Required, but never shown












            Required, but never shown







            Required, but never shown







            Popular posts from this blog

            Plaza Victoria

            Puebla de Zaragoza

            Musa