
{"id":9095859,"date":"2016-04-25T15:29:19","date_gmt":"2016-04-26T01:29:19","guid":{"rendered":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/?p=9095859"},"modified":"2016-04-25T15:35:02","modified_gmt":"2016-04-26T01:35:02","slug":"the-tukey-test-with-known-variance","status":"publish","type":"post","link":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/the-tukey-test-with-known-variance\/","title":{"rendered":"The Tukey test: starting with a known variance"},"content":{"rendered":"<p>Consider normal random variables $Y_i$ with means $\\mu_i$, $1\\le i\\le k$.<br \/>\nThe Tukey test often seems overly complicated, but it becomes clearer if we first assume that variance of $Y_i$, $\\sigma^2$, is known.<\/p>\n<p>Let&#8217;s draw $r$ samples from each, called $Y_{i,j}$, $1\\le j\\le r$.<br \/>\nDenote the sample averages by $\\overline Y_{i+}$.<br \/>\nLet $W_i = \\overline Y_{i+}-\\mu_i$.<br \/>\nLet $Q=R$, the range, be defined by<br \/>\n$$<br \/>\nR = \\max_i W_i &#8211; \\min_i W_i = \\max_{i,j} |W_i-W_j|.<br \/>\n$$<br \/>\nSuppose we understand the distribution of $R$ well enough to find a number $Q_\\alpha$ such that<br \/>\n$$<br \/>\n\\Pr(R\\le Q_\\alpha) = 1-\\alpha<br \/>\n$$<br \/>\nDefine the intervals<br \/>\n$$<br \/>\nI_{i,j} = (\\overline Y_{i+}-\\overline Y_{j+} &#8211; Q_\\alpha, \\overline Y_{i+}-\\overline Y_{j+} + Q_\\alpha)<br \/>\n$$<br \/>\nThen it is easily seen that<br \/>\n$$<br \/>\n\\Pr(\\text{for all $i$, $j$, }\\mu_i-\\mu_j \\in I_{i,j}) = 1-\\alpha.<br \/>\n$$<br \/>\nand hence for all $i$, $j$,<br \/>\n$$<br \/>\n\\Pr(\\mu_i-\\mu_j\\in I_{i,j}) \\ge 1-\\alpha.<br \/>\n$$<br \/>\n(Larsen and Marx make a mistake here, mixing up the last two equations.)<br \/>\nThus, the hypothesis that $\\mu_i=\\mu_j$ can be rejected if we observe $0\\not\\in I_{i,j}$.<\/p>\n<p>Now, if the variance is unknown, we instead define $Q=R\/S$ where $S$ is a certain estimator of the variance.<br \/>\nAn important point is that we want to understand the joint distribution of $R$ and $S$, which is easiest if they are independent.<br \/>\nWe do have an estimator of $\\sigma^2$ that&#8217;s independent of the $W_i$, namely the residual sum of squares (a.k.a. sum of squares for error),<br \/>\n$$<br \/>\n\\mathrm{SSE} = \\sum_i \\sum_j (Y_{i,j}-\\overline Y_{i+})^2<br \/>\n$$<br \/>\nSo we take $S^2$ to be a suitable constant time $\\mathrm{SSE}$.<br \/>\nNamely, we want $S^2$ to be an unbiased estimator of $\\sigma^2\/r$, the variance of $W_i$. And we know that $\\mathrm{SSE}\/\\sigma^2$ is $\\chi^2(rk-k)$ distributed.<br \/>\nThis leads us to define<br \/>\n$$<br \/>\nS^2 = \\mathrm{MSE}\/r<br \/>\n$$<br \/>\nwhere $\\mathrm{MSE} = \\mathrm{SSE}\/(rk-k)$.<br \/>\nA point here is that $\\mathrm{SSE}\/\\sigma^2$ is a sum of $k$ independent $\\chi^2(r-1)$ random variables (one for each $i$), hence is itself $\\chi^2(rk-k)$.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Consider normal random variables $Y_i$ with means $\\mu_i$, $1\\le i\\le k$. The Tukey test often seems overly complicated, but it becomes clearer if we first assume that variance of $Y_i$, $\\sigma^2$, is known. Let&#8217;s draw $r$ samples from each, called $Y_{i,j}$, $1\\le j\\le r$. Denote the sample averages by $\\overline Y_{i+}$. Let $W_i = \\overline &hellip; <a href=\"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/the-tukey-test-with-known-variance\/\" class=\"more-link\">Continue reading <span class=\"screen-reader-text\">The Tukey test: starting with a known variance<\/span> <span class=\"meta-nav\">&rarr;<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[4],"tags":[],"class_list":["post-9095859","post","type-post","status-publish","format-standard","hentry","category-teaching"],"_links":{"self":[{"href":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/wp-json\/wp\/v2\/posts\/9095859","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/wp-json\/wp\/v2\/comments?post=9095859"}],"version-history":[{"count":10,"href":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/wp-json\/wp\/v2\/posts\/9095859\/revisions"}],"predecessor-version":[{"id":9095869,"href":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/wp-json\/wp\/v2\/posts\/9095859\/revisions\/9095869"}],"wp:attachment":[{"href":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/wp-json\/wp\/v2\/media?parent=9095859"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/wp-json\/wp\/v2\/categories?post=9095859"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/math.hawaii.edu\/wordpress\/bjoern\/wp-json\/wp\/v2\/tags?post=9095859"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}