Clarifications-requests

[francescoveri]

Hi Dave,
just to clarify and add:

1. Intersubjective correlation in each pair for Q and P is a Spearman rho value? I have slightly different results when I am employing SPSS or R.
2. is it possible to add the orthogonal distance of each pairs d=(∣x-y∣)/√2 between Intersubjective correlation Q and P?
3. then we should calculate the arithmetic mean of such distances D=1/n ∑d for each stage of the deliberation process
4. then normalize the distance as follows:
   a) IC=1-D;
   b) IC_min=(1-(√2/2))
   b) IC_norm=2*((IC-IC_min)/(1-IC_min))-1
   Please let me know whether you have any question in relation to the above

[davidmoten]

Hi Francesco,

1. Spearman's is used in the latest version as the default correlation coefficient. The code used in its calculation is https://github.com/DeliberativeAnalysis/DeliberateQ/blob/master/src/main/java/com/github/deliberateq/util/math/CorrelationCoefficient.java and unit tests are in https://github.com/DeliberativeAnalysis/DeliberateQ/blob/master/src/test/java/com/github/deliberateq/util/math/CorrelationCoefficientTest.java.

The unit test used is this:

@Test
    public void testSpearmansCorrelation() {
        Vector a = new Vector(56, 75, 45, 71, 62, 64, 58, 80, 76, 61);
        Vector b = new Vector(66, 70, 40, 60, 65, 56, 59, 77, 67, 63);
        assertEquals(1 - 324.0 / 990.0, CorrelationCoefficient.SPEARMANS.apply(a, b), 0.0000001);
    }

If you disagree with the expectation of this unit test then let me know and we'll investigate!

[francescoveri]

Yes, it is correct. Thanks for the clarification FV Francesco Veri Postdoctoral Research Associate Centre for Deliberative Democrac​y & Global Governance The Institute for Governance and Policy Analysis Building 24 | University of Canberra | ACT 2617, Australia

…

-------------------------- Email: [email protected] ​ Most recent publications: Veri, Francesco. (2019). "Explaining Foreigners' Political Rights in the Context of Direct Democracy: A Fuzzy-Set QCA if Swiss Cantonal Popular Votes<https://www.cogitatiopress.com/politicsandgovernance/article/view/1779>". Politics and Governance. 7(2): 410-426. Veri, Francesco. (2019). "Aggregation bias and ambivalent cases: a new parameter of consistency in fuzzy-set Qualitative Comparative Analysis (fsQCA)<https://brill.com/abstract/journals/coso/18/2/article-p229_5.xml?rskey=rPZ8NT&result=1>". Comparative Sociology. 18(2): 229-255.

________________________________ From: Dave Moten <[email protected]> Sent: Wednesday, 22 July 2020 9:40 AM To: DeliberativeAnalysis/DeliberateQ <[email protected]> Cc: Francesco.Veri <[email protected]>; Author <[email protected]> Subject: Re: [DeliberativeAnalysis/DeliberateQ] Clarifications-requests (#14) Hi Francesco, 1. Spearman's is used in the latest version as the default correlation coefficient. The code used in its calculation is https://github.com/DeliberativeAnalysis/DeliberateQ/blob/master/src/main/java/com/github/deliberateq/util/math/CorrelationCoefficient.java<https://github.com/DeliberativeAnalysis/DeliberateQ/blob/master/src/main/java/com/github/deliberateq/util/math/CorrelationCoefficient.java> and unit tests are in https://github.com/DeliberativeAnalysis/DeliberateQ/blob/master/src/test/java/com/github/deliberateq/util/math/CorrelationCoefficientTest.java<https://github.com/DeliberativeAnalysis/DeliberateQ/blob/master/src/test/java/com/github/deliberateq/util/math/CorrelationCoefficientTest.java>. The unit test used is this: @test public void testSpearmansCorrelation() { Vector a = new Vector(56, 75, 45, 71, 62, 64, 58, 80, 76, 61); Vector b = new Vector(66, 70, 40, 60, 65, 56, 59, 77, 67, 63); assertEquals(1 - 324.0 / 990.0, CorrelationCoefficient.SPEARMANS.apply(a, b), 0.0000001); } If you disagree with the expectation of this unit test then let me know and we'll investigate! — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#14 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AQFAKIPQ4YP63ZL4FMRJO5LR4YRNZANCNFSM4PDBLZOQ>.

[davidmoten]

Re 2-4, we can add whatever calculation you like. Where do you want to see it used or displayed?

[davidmoten]

Hey Franceso, got an answer for my last question?

[francescoveri]

Hi Dave,
Sorry I am in holiday and I forgot to reply. if it is possible the orthogonal distance should appear when you scroll the mouse over each point and add a column into the tables with the correlations.
The arithmetic mean of such distances should be added in the box analyse (with the PCA) and work in the same way of PCA now (i.e., display the result under first stage, and third stage).
Both measures should be add in First stage, third stage of deliberation process.
Have a good weekend
Francesco

[davidmoten]

Excellent excuse Franceso, enjoy your break!

…

On Sat, 1 Aug 2020 at 09:24, francescoveri ***@***.***> wrote: Hi Dave, Sorry I am in holiday and I forgot to reply. if it is possible the orthogonal distance should appear when you scroll the mouse over each point and add a column into the tables with the correlations. The arithmetic mean of such distances should be added in the box analyse (with the PCA) and work in the same way of PCA now (i.e., display the result under first stage, and third stage). Both measures should be add in First stage, third stage of deliberation process. Have a good weekend Francesco — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#14 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACNV2ZXF5VTSMKMDEWALYDR6NHA7ANCNFSM4PDBLZOQ> .