Subject: analysis of paired comparison data From: Delphine Devallez <delphine.devallez@xxxxxxxx> Date: Tue, 24 Feb 2009 16:25:24 +0100 List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>--000e0cd28e9c5f93600463abbaa1 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Dear all, I recently carried out a listening test on the perception of auditory distance using paired comparison judgments. To construct a psychological scale, I tried to fit a Thurstone-Mosteller Case V model as well as a Bradley-Terry-Luce model (which give similar results). Unfortunately for both models the goodness of fit (chi-square test) gives a very low p-value, meaning that the models should be rejected. Then, I was thinking about trying to apply the Thurstone-Mosteller Case III (which does not assume that the discriminal dispersions are equal), however I don't know how to do it (in particular how to estimate the dispersion values), and I did not find any literature on this case. Does anyone have some experience with such analysis? Thank you in advance! Kind Regards, Delphine Devallez ************************************************ Address: Dipartimento di Informatica Universit=E0 degli Studi di Verona Ca' Vignal 2, Strada Le Grazie 15 37134 Verona, Italy Phone: +39 045 802 7968 (office) E-mail: delphine.devallez@xxxxxxxx --000e0cd28e9c5f93600463abbaa1 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Dear all,<br><div class=3D"gmail_quote"><br>I recently carried out a listen= ing test on the perception of auditory distance using paired comparison jud= gments. To construct a psychological scale, I tried to fit a Thurstone-Most= eller Case V model as well as a Bradley-Terry-Luce model (which give simila= r results). Unfortunately for both models the goodness of fit (chi-square t= est) gives a very low p-value, meaning that the models should be rejected. = <br> Then, I was thinking about trying to apply the Thurstone-Mosteller Case III= (which does not assume that the discriminal dispersions are equal), howeve= r I don't know how to do it (in particular how to estimate the dispersi= on values), and I did not find any literature on this case. Does anyone hav= e some experience with such analysis?<br> <br>Thank you in advance!<br clear=3D"all"><br>Kind Regards,<br><font color= =3D"#888888"><br><br></font></div><br>Delphine Devallez<br>****************= ********************************<br>Address: =A0Dipartimento di Informatica= <br> Universit=E0 degli Studi di Verona<br> Ca' Vignal 2, Strada Le Gra= zie 15<br> 37134 Verona, Italy<br>Phone: +39 045 802 7968 (office)<br>= E-mail: <a href=3D"mailto:delphine.devallez@xxxxxxxx">delphine.devallez@xxxxxxxx= mail.com</a><br> --000e0cd28e9c5f93600463abbaa1--