{"id":51,"date":"2011-05-06T08:10:16","date_gmt":"2011-05-06T16:10:16","guid":{"rendered":"https:\/\/blogs.ubc.ca\/poli574\/?p=51"},"modified":"2011-05-12T08:40:29","modified_gmt":"2011-05-12T16:40:29","slug":"discrete-choice-multiple-categorical-outcomes-ldv","status":"publish","type":"post","link":"https:\/\/blogs.ubc.ca\/poli574\/2011\/05\/06\/discrete-choice-multiple-categorical-outcomes-ldv\/","title":{"rendered":"Discrete Choice &#8211; Multiple Categorical Outcomes &#8211; LDV"},"content":{"rendered":"<p><span style=\"font-family: Calibri;\"> <\/span><\/p>\n<p><span style=\"font-family: Calibri;\"> <\/span><\/p>\n<p><span style=\"font-size: medium;\"> <\/span><\/p>\n<h2><strong>Introduction <\/strong><\/h2>\n<p>This module deals with statistical models to help us explain <em>multiple, unordered, categorical outcomes<\/em>.<\/p>\n<p>It should help you learn, relearn, and do a number of things:<\/p>\n<ul>\n<li>Foundations of maximum likelihood estimation<\/li>\n<li>Practicalities of maximum likelihood estimation and working with likelihoods (a bit)<\/li>\n<li>Using Clarify and Margins to interpret non-linear models (logit, and beyond)<\/li>\n<li>Understand, use, and interpret\n<ul>\n<li>Multinomial logit<\/li>\n<li>Multinomial probit<\/li>\n<li>Conditional logit<\/li>\n<li>Selection models<\/li>\n<\/ul>\n<\/li>\n<li>More stata tricks<\/li>\n<\/ul>\n<p>Here are my teaching notes: <a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/574-Cutler-teaching-notes-Revised.pdf\">574 Cutler teaching notes Revised<\/a><\/p>\n<h2><strong><span>1. Maximum-Likelihood Estimation &amp; Logit refresher<\/span><\/strong><\/h2>\n<p>First, you should make sure you&#8217;re on board with maximum likelihood estimation. Logit, for example.\u00a0And interpreting effects of a logit in Stata with Clarify and\/or the new &#8220;margins&#8221; command.<\/p>\n<p>One way to review MLE in general is to watch the screen and listen to me as I did the relevant class for this year&#8217;s POLI572. It\u2019s slow moving, however, so it may not be the most efficient thing to do.<\/p>\n<p><a href=\"http:\/\/www.screencast.com\/t\/kYG2I1ETE\">http:\/\/www.screencast.com\/t\/kYG2I1ETE<\/a><\/p>\n<p>[You actually have to start in the middle because I cut it into two parts and then when I reassembled it I put the two parts backwards. So you have to wait for a bunch of it to download, start at 1:05:00 and follow it to the end, and then for the second half of the class, just start at the very beginning. The very beginning stuff is post-break, so it&#8217;s very silly for a few minutes &#8212; you can restart the second half at about 4:00.]<\/p>\n<p>Another way to refresh is for you to consult the materials listed and linked from the POLI572 blog: <a href=\"https:\/\/blogs.ubc.ca\/poli5722011\/\">https:\/\/blogs.ubc.ca\/poli5722011\/<\/a> under &#8220;FOR MARCH 18&#8221;.<\/p>\n<p>The quickest and simplest is the Kellstedt and Whitten text: section 11.4 for a quick logit refresher.<\/p>\n<p>I strongly recommend the following new reading that will both reinforce what logit is doing and actually help you think about using interactions in logit\/probit models is:<\/p>\n<p>William D. Berry,\u00a0Jacqueline H. R. DeMeritt,\u00a0Justin Esarey. 2009. <strong>Testing for Interaction in Binary Logit and Probit Models: Is a Product Term Essential? <\/strong><em>AJPS. <\/em>2009 <a href=\"http:\/\/onlinelibrary.wiley.com\/doi\/10.1111\/j.1540-5907.2009.00429.x\/pdf\">DOI:\u00a010.1111\/j.1540-5907.2009.00429.<\/a><\/p>\n<h2><strong><a href=\"http:\/\/onlinelibrary.wiley.com\/doi\/10.1111\/j.1540-5907.2009.00429.x\/pdf\"><\/a>2. Multiple Categorical Outcomes<\/strong><\/h2>\n<p>Then, SECOND, we can move on to starting to think about MULTIPLE unordered outcomes.<\/p>\n<p>The most obvious place this applies in our discipline is in vote choice in multiparty systems. I want to emphasize, however, that there are lots of, and probably more important, questions across subfields involving multiple unordered (discrete) outcomes.<\/p>\n<p>Nevertheless, because I know this stuff best in the context of voting research and because I have cool, brand new CES data, and because I have actually published using a conditional logit model, I\u2019ll stick with this as an example.\u00a0 We\u2019ll try to explain vote choice in Quebec in this very recent election. \u00a0That&#8217;ll be the data you can play with. It will appear here with a do file on Monday.<\/p>\n<p>Oh, Tuesday officially. Here&#8217;s a zip file with data and do file: <a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/574-class-1.zip\">574-class 1<\/a><\/p>\n<p>I&#8217;ve got three levels of &#8216;textbook&#8217; for this module.<\/p>\n<p>From most detailed and hardest to follow to least:<br \/>\n(So you may want to start with the bottom one if you want to ease into it; the top one if you&#8217;re confident you can follow the math. I would like you to have the Golder notes printed or otherwise handy for reference).<\/p>\n<ol>\n<li>Matt Golder&#8217;s teaching notes (<a href=\"http:\/\/homepages.nyu.edu\/~mrg217\/unordered_multiresponse.pdf\">http:\/\/homepages.nyu.edu\/~mrg217\/unordered_multiresponse.pdf<\/a>)<br \/>\nNOTE that Conditional logit (3) precedes Multinomial Logit (4) in Golder&#8217;s notes.<\/li>\n<li><a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/Kousser-Unordered-Choice-Models.pdf\">Kousser Unordered Choice Models<\/a> (pdf)<\/li>\n<li>and, for a more basic summary with some Stata-ish-ness: <a href=\"http:\/\/www.nd.edu\/~rwilliam\/stats2\/l92.pdf\">http:\/\/www.nd.edu\/~rwilliam\/stats2\/l92.pdf<\/a><\/li>\n<\/ol>\n<p>Then (or, first):<\/p>\n<p><strong><a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/Tucker-et-al-Transitional-Winners-and-Losers-AJPS-2002.pdf\">Tucker et al Transitional Winners and Losers AJPS 2002<\/a><\/strong><\/p>\n<p><strong>Transitional Winners and Losers: Attitudes toward EU Membership in Post-Communist Countries, <\/strong>Joshua A. Tucker, Alexander C. Pacek and Adam J. Berinsky. \u00a0<em>American Journal of Political Science, <\/em>Vol. 46, No. 3 (Jul., 2002), pp. 557-571\u00a0Article Stable URL:\u00a0<a href=\"http:\/\/www.jstor.org\/stable\/3088399\">http:\/\/www.jstor.org\/stable\/3088399<\/a><\/p>\n<ul>\n<li>This is a good application of Multinomial Logit.<\/li>\n<\/ul>\n<p><strong><a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/Alvarez-Nagler-Models-Collide-AJPS-1998.pdf\">Alvarez Nagler Models Collide AJPS 1998<\/a><\/strong><\/p>\n<p><strong><a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/Alvarez-Nagler-Models-Collide-AJPS-1998.pdf\"><\/a>When Politics and Models Collide: Estimating Models of Multiparty Elections<br \/>\n<\/strong>R. Michael Alvarez and Jonathan Nagler,\u00a0<em>American Journal of Political Science, <\/em>Vol. 42, No. 1 (Jan., 1998), pp. 55-96\u00a0Article Stable URL: <a href=\"http:\/\/www.jstor.org\/stable\/2991747\">http:\/\/www.jstor.org\/stable\/2991747<\/a><\/p>\n<ul>\n<li>This shows that MNL is equivalent to binary logits, but slightly more efficient.<\/li>\n<\/ul>\n<p><a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/Dow-Endersby-MNP-vs.-MNL-JELS-2004.pdf\"><strong>Dow Endersby MNP vs. MNL JELS 2004<\/strong><\/a><\/p>\n<p><strong>Multinomial probit and multinomial logit: a comparison of choice models for voting research. <\/strong>Dow, Jay K. and James W. Endersby.\u00a0<a href=\"http:\/\/www.sciencedirect.com\/science\/journal\/02613794\">Electoral Studies<\/a> <a href=\"http:\/\/www.sciencedirect.com\/science?_ob=PublicationURL&amp;_tockey=%23TOC%235904%232004%23999769998%23475179%23FLA%23&amp;_cdi=5904&amp;_pubType=J&amp;view=c&amp;_auth=y&amp;_acct=C000050484&amp;_version=1&amp;_urlVersion=0&amp;_userid=1022551&amp;md5=2a2f2e8afec10250dbf60afff67f8b61\">Volume 23, Issue 1<\/a>, March 2004, Pages 107-122 \u00a0<a href=\"http:\/\/dx.doi.org\/10.1016\/S0261-3794(03)00040-4\">doi:10.1016\/S0261-3794(03)00040-4<\/a><\/p>\n<ul>\n<li>This shows that MNP is kind-of a waste of time<\/li>\n<\/ul>\n<p><a href=\"http:\/\/homepages.nyu.edu\/~mrg217\/unordered_multiresponse2.pdf\">Golder&#8217;s Discrete Choice Notes II<\/a><\/p>\n<ul>\n<li>This is a mini-textbook for Nested logit and Multinomial Probit<\/li>\n<\/ul>\n<h2>3. Characteristics of CHOICES, not just of choosers: Conditional Logit<\/h2>\n<p>See the Golder notes, above, for the math of the conditional logit model.<\/p>\n<p>See this <a href=\"http:\/\/data.princeton.edu\/wws509\/stata\/c6s3.html\">Princeton page<\/a> for the Stata operationalization of the model.<\/p>\n<p><strong><a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/Cutler-Simplest-Shortcut-of-All-JOP-2001.pdf\">Cutler Simplest Shortcut of All JOP 2001<\/a><\/strong><\/p>\n<p><strong><a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/Cutler-Simplest-Shortcut-of-All-JOP-2001.pdf\"><\/a>The Simplest Shortcut of All: Sociodemographic Characteristics and Electoral Choice. <\/strong>Fred Cutler. 2002.\u00a0<em>The Journal of Politics. <\/em>Vol. 64, No. 2, pp. 466-490.\u00a0Article Stable URL: <a href=\"http:\/\/www.jstor.org\/stable\/2691857\">http:\/\/www.jstor.org\/stable\/2691857<\/a><\/p>\n<h2>4. Selection models<\/h2>\n<p>We may make it to selection models. I&#8217;d like to.<\/p>\n<p>Start with Wooldridge section 17.5.<\/p>\n<p>Or, maybe there&#8217;s no point. Again, Matt Golder&#8217;s notes are excellent: <a href=\"http:\/\/homepages.nyu.edu\/~mrg217\/selection.pdf\">http:\/\/homepages.nyu.edu\/~mrg217\/selection.pdf<\/a><\/p>\n<p>Then, for some examples, which you can read according to your interests<\/p>\n<p><a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/Berinsky-two-faces.pdf\">Berinsky two faces of public opinion<\/a><\/p>\n<p><a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/Blanton-Sanctions-AJPS.pdf\">Blanton Sanctions AJPS<\/a><\/p>\n<p><a href=\"https:\/\/blogs.ubc.ca\/poli574\/files\/2011\/05\/Nooruddin-sanctions1.pdf\">Nooruddin sanctions[1]<\/a><\/p>\n<p><strong>A Summary of Political Selection: The Effect of Strategic Choice on the Escalation of International Crises<\/strong><\/p>\n<p>Alastair Smith,\u00a0<em>American Journal of Political Science. <\/em>Vol. 42, No. 2 (Apr., 1998), pp. 698-701.\u00a0Article Stable URL: <a href=\"http:\/\/www.jstor.org\/stable\/2991776\">http:\/\/www.jstor.org\/stable\/2991776<\/a><\/p>\n<p>Sorry, that one&#8217;s just a summary of a poster. This one is better:<\/p>\n<div id=\"_mcePaste\"><strong>A Unified Statistical Model of Conflict Onset and Escalation<\/strong><\/div>\n<p>William Reed<br \/>\nAmerican Journal of Political Science, Vol. 44, No. 1 (Jan., 2000), pp. 84-93<br \/>\nStable URL: <a href=\"http:\/\/www.jstor.org\/stable\/2669294\">http:\/\/www.jstor.org\/stable\/2669294<\/a><\/p>\n<p>James Raymond Vreeland (2008). <a href=\"http:\/\/journals.cambridge.org\/action\/displayFulltext?type=1&amp;fid=1630360&amp;jid=INO&amp;volumeId=62&amp;issueId=01&amp;aid=1630356&amp;bodyId=&amp;membershipNumber=&amp;societyETOCSession=\">Political Institutions and Human Rights: Why Dictatorships Enter into the United Nations Convention Against Torture.<\/a> International Organization, 62, pp 65-101<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Introduction This module deals with statistical models to help us explain multiple, unordered, categorical outcomes. It should help you learn, relearn, and do a number of things: Foundations of maximum likelihood estimation Practicalities of maximum likelihood estimation and working with likelihoods (a bit) Using Clarify and Margins to interpret non-linear models (logit, and beyond) Understand, [&hellip;]<\/p>\n","protected":false},"author":1822,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-51","post","type-post","status-publish","format-standard","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/blogs.ubc.ca\/poli574\/wp-json\/wp\/v2\/posts\/51","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blogs.ubc.ca\/poli574\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blogs.ubc.ca\/poli574\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blogs.ubc.ca\/poli574\/wp-json\/wp\/v2\/users\/1822"}],"replies":[{"embeddable":true,"href":"https:\/\/blogs.ubc.ca\/poli574\/wp-json\/wp\/v2\/comments?post=51"}],"version-history":[{"count":21,"href":"https:\/\/blogs.ubc.ca\/poli574\/wp-json\/wp\/v2\/posts\/51\/revisions"}],"predecessor-version":[{"id":55,"href":"https:\/\/blogs.ubc.ca\/poli574\/wp-json\/wp\/v2\/posts\/51\/revisions\/55"}],"wp:attachment":[{"href":"https:\/\/blogs.ubc.ca\/poli574\/wp-json\/wp\/v2\/media?parent=51"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blogs.ubc.ca\/poli574\/wp-json\/wp\/v2\/categories?post=51"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blogs.ubc.ca\/poli574\/wp-json\/wp\/v2\/tags?post=51"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}