<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Statistics on 好豪筆記</title><link>https://haosquare.com/tags/statistics/</link><description>Recent content in Statistics on 好豪筆記</description><generator>Hugo -- gohugo.io</generator><language>zh-tw</language><managingEditor>haosquare.tw@gmail.com (好豪)</managingEditor><webMaster>haosquare.tw@gmail.com (好豪)</webMaster><copyright>2026 好豪筆記 ʕ •ᴥ•ʔ</copyright><lastBuildDate>Sat, 01 Nov 2025 00:00:00 +0000</lastBuildDate><atom:link href="https://haosquare.com/tags/statistics/index.xml" rel="self" type="application/rss+xml"/><item><title>Sequential A/B Testing：讓你「偷看」實驗、快速決策的神奇武器</title><link>https://haosquare.com/sequential-ab-testing/</link><pubDate>Sat, 01 Nov 2025 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/sequential-ab-testing/</guid><description>Sequential A/B Testing 不止允許偷看數據、需要的樣本數比標準假設檢定還少！白話介紹這個 Netflix、Spotify 都在用的神奇方法</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/sequential-ab-testing/featured.jpg"/></item><item><title>A/B 測試觸發機制實戰指南：Counterfactual Logging 讓機器學習演算法的實驗超高效率！</title><link>https://haosquare.com/ab-testing-trigger-counterfactual-logging/</link><pubDate>Fri, 10 Oct 2025 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/ab-testing-trigger-counterfactual-logging/</guid><description>深入解析三種 A/B 測試觸發方法：使用者層級觸發、Exposure Logging、Counterfactual Logging，搭配 DoorDash、Spotify 等實際業界案例，幫助你解決實驗稀釋效果問題</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/ab-testing-trigger-counterfactual-logging/featured.jpg"/></item><item><title>A/B 測試的觸發（Trigger）：為什麼你需要精準定位受影響使用者？</title><link>https://haosquare.com/ab-testing-trigger/</link><pubDate>Fri, 19 Sep 2025 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/ab-testing-trigger/</guid><description>介紹 A/B 測試觸發機制（Trigger）如何精準定位受實驗影響的使用者，透過排除雜訊提升統計檢定力，讓所需樣本數減少一半、大大加速實驗</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/ab-testing-trigger/featured.jpg"/></item><item><title>P 值越小越好？破除新手誤解：你需要信賴區間與效果量</title><link>https://haosquare.com/p-value-smaller-better-myth/</link><pubDate>Tue, 27 May 2025 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/p-value-smaller-better-myth/</guid><description>P 值越小代表結果越重要？統計顯著與「實用」顯著是兩回事！</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/p-value-smaller-better-myth/featured.webp"/></item><item><title>t 檢定與中央極限定理：你真的懂常態分佈嗎？</title><link>https://haosquare.com/normal-distribution-central-limit-theorem-t-test/</link><pubDate>Sun, 10 Nov 2024 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/normal-distribution-central-limit-theorem-t-test/</guid><description>常態分佈是一個非常基礎但又容易被誤解的概念，只要把各個知識點連結起來，其實沒那麼複雜</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/normal-distribution-central-limit-theorem-t-test/featured.png"/></item><item><title>Naive Bayes 分類模型：簡單卻強大的機器學習基本功 — Udemy 課程心得</title><link>https://haosquare.com/udemy-naive-bayes-intro/</link><pubDate>Sun, 14 Jul 2024 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/udemy-naive-bayes-intro/</guid><description>儘管深度學習模型光彩奪目，Naive Bayes 依然因為它簡單、高效、好解釋而有其地位</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/udemy-naive-bayes-intro/featured.jpg"/></item><item><title>A/B Test 看到顯著也不能信？小心統計檢定力不足！</title><link>https://haosquare.com/ab-test-power-analysis-intuition-busters/</link><pubDate>Sun, 26 May 2024 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/ab-test-power-analysis-intuition-busters/</guid><description>統計檢定力 Power Analysis</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/ab-test-power-analysis-intuition-busters/featured.jpg"/></item><item><title>A/B Test 如何設定顯著水準？從 0.05 改成 0.01 算是夠嚴格的實驗嗎？</title><link>https://haosquare.com/ab-test-significance-level-false-positive-rate/</link><pubDate>Sun, 12 May 2024 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/ab-test-significance-level-false-positive-rate/</guid><description>介紹偽陽性率，它與顯著水準、統計檢定力、以及決策品質的關聯</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/ab-test-significance-level-false-positive-rate/featured.jpg"/></item><item><title>倖存者偏差，怎麼在數據分析中避免？居然還有死者偏誤！</title><link>https://haosquare.com/survivorship-bias-and-right-censored-data/</link><pubDate>Mon, 22 Apr 2024 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/survivorship-bias-and-right-censored-data/</guid><description>分享極其常見的倖存者偏差、以及與其相似的「死者偏誤」與右設限資料</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/survivorship-bias-and-right-censored-data/featured.jpg"/></item><item><title>卡方檢定的三大實戰使用時機：A/B 測試、分類模型、特徵選擇</title><link>https://haosquare.com/chi-squared-test-application/</link><pubDate>Sun, 14 Apr 2024 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/chi-squared-test-application/</guid><description>資料科學家不得不知道的三大卡方檢定使用時機：多組別的 A/B 測試、評估二分類模型、以及機器學習模型特徵選擇</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/chi-squared-test-application/featured.jpg"/></item><item><title>棒打出頭鳥：L1 與 L2 正規化防止模型中的係數過大</title><link>https://haosquare.com/l1-l2-regularization/</link><pubDate>Fri, 15 Mar 2024 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/l1-l2-regularization/</guid><description>L1 與 L2 正規化（Regularization）透過避免模型中的係數過大，減緩機器學習模型的過度擬合（Overfitting）問題</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/l1-l2-regularization/featured.jpg"/></item><item><title>型一與型二錯誤介紹，圖解快速學習！</title><link>https://haosquare.com/type-i-error-and-type-ii-error/</link><pubDate>Sat, 09 Dec 2023 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/type-i-error-and-type-ii-error/</guid><description>如果你還無法搞懂 P 值、顯著水準、或檢定力，型一與型二錯誤正是它們共通的重要觀念，這篇文章將用圖示讓你輕鬆理解！</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/type-i-error-and-type-ii-error/featured.jpg"/></item><item><title>卡方檢定一篇看懂，白話介紹三種檢定方式</title><link>https://haosquare.com/chi-squared-test/</link><pubDate>Wed, 04 Oct 2023 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/chi-squared-test/</guid><description>白話介紹三種最常見的卡方檢定：配適度檢定、獨立性檢定、同質性檢定，教學如何用 Python 的 Scipy 套件來實作</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/chi-squared-test/featured.jpg"/></item><item><title>ROC 曲線上的最佳閾值：Youden Index 與圖解法介紹</title><link>https://haosquare.com/roc-curve-best-cutoff/</link><pubDate>Sat, 23 Sep 2023 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/roc-curve-best-cutoff/</guid><description>3 種在 ROC 曲線上選擇最佳閾值的方法，包括 Youden&amp;rsquo;s J statistic、最近距離法、還有一種用肉眼就能看出最佳閾值的平行線判別法</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/roc-curve-best-cutoff/featured.jpg"/></item><item><title>圖解 ROC 曲線：精通 ROC 與 AUC 用法、輕鬆記熟定義</title><link>https://haosquare.com/roc-curve/</link><pubDate>Fri, 15 Sep 2023 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/roc-curve/</guid><description>讀完這篇將再也不會忘記 ROC 曲線、TPR、FPR 的定義！</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/roc-curve/featured.png"/></item><item><title>A/B Test 樣本數究竟該怎麼算？公式與 Python 算法通通學會</title><link>https://haosquare.com/calculate-ab-testing-sample-size/</link><pubDate>Tue, 27 Jun 2023 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/calculate-ab-testing-sample-size/</guid><description>A/B Test 樣本數計算的公式與假設</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/calculate-ab-testing-sample-size/featured.jpg"/></item><item><title>單尾與雙尾檢定，該選哪一個？</title><link>https://haosquare.com/one-tailed-vs-two-tailed-test/</link><pubDate>Sun, 05 Mar 2023 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/one-tailed-vs-two-tailed-test/</guid><description>簡短說明單尾與雙尾檢定的統計學特性差異，也介紹單尾檢定常見的錯誤使用方法</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/one-tailed-vs-two-tailed-test/featured.jpg"/></item><item><title>為何要用貝氏統計？讓決策持續進化的灰階思考</title><link>https://haosquare.com/why-bayeisan-statistics/</link><pubDate>Sat, 26 Nov 2022 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/why-bayeisan-statistics/</guid><description>貝氏統計幫助你在模型加入個人觀點、讓統計模型隨著新的資料取得不斷進化、並且量化你對決策究竟有多麽「不確定」</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/why-bayeisan-statistics/featured.png"/></item><item><title>Novelty 與 Primacy Effect：好奇心殺死 A/B Test？</title><link>https://haosquare.com/ab-testing-novelty-primacy-effect/</link><pubDate>Sun, 03 Jul 2022 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/ab-testing-novelty-primacy-effect/</guid><description>A/B 測試不得不知道的 Novelty 與 Primacy Effect 問題，了解顧客的「好奇」與「守舊」傾向影響 A/B Testing 判讀</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/ab-testing-novelty-primacy-effect/featured.jpg"/></item><item><title>四個步驟徹底學習標準誤：它跟標準差到底有什麼不同？</title><link>https://haosquare.com/standard-error/</link><pubDate>Sun, 15 May 2022 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/standard-error/</guid><description>標準誤與標準差有什麼差別？</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/standard-error/featured.jpg"/></item><item><title>因果推論簡介：A/B Testing 行不通時怎麼辦？</title><link>https://haosquare.com/causal-inference-intro/</link><pubDate>Sun, 13 Mar 2022 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/causal-inference-intro/</guid><description>簡介因果推論（Causal Inference）這個研究領域，用範例說明分析因果關係的三大困難點：混淆變項、自我選擇偏誤、以及反事實</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/causal-inference-intro/featured.jpg"/></item><item><title>從《天堂 M》手遊機率爭議學會二項式檢定與 P-value 解讀</title><link>https://haosquare.com/lineagem-binomial-test-p-value/</link><pubDate>Sat, 12 Feb 2022 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/lineagem-binomial-test-p-value/</guid><description>2介紹二項式檢定如何檢驗抽獎成功率，並且帶你一步步計算 P-value、學會清楚解讀 P-value</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/lineagem-binomial-test-p-value/featured.jpg"/></item><item><title>圖解信賴區間，兩種常見的錯誤解讀</title><link>https://haosquare.com/confidence-interval/</link><pubDate>Thu, 20 Jan 2022 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/confidence-interval/</guid><description>你知道信賴區間的 95% 機率是指什麼嗎？標準差跟標準誤哪裡不一樣？</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/confidence-interval/featured.jpg"/></item><item><title>多重檢定問題：一定會有人中樂透</title><link>https://haosquare.com/multiple-testing-intro/</link><pubDate>Mon, 06 Dec 2021 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/multiple-testing-intro/</guid><description>分享什麼是多重檢定問題，他到底跟樂透、A/B Test、或者探索性資料分析有什麼關聯，並且介紹用來解決多重檢定問題的 Bonferroni 校正</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/multiple-testing-intro/featured.jpg"/></item><item><title>比例資料的 A/B Test 該用卡方還是 Z 檢定？</title><link>https://haosquare.com/hypothesis-testing-for-proportion/</link><pubDate>Fri, 06 Aug 2021 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/hypothesis-testing-for-proportion/</guid><description>介紹 Z 檢定與卡方檢定之間的關聯，以及什麼情況下這兩者都可以使用</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/hypothesis-testing-for-proportion/featured.jpg"/></item><item><title>A/A Test：商業實驗不能忘的前置動作</title><link>https://haosquare.com/aa-testing-before-ab-testing/</link><pubDate>Sat, 17 Jul 2021 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/aa-testing-before-ab-testing/</guid><description>A/A Test 跟 A/B Test 不一樣，A/A Test 是 A/B Test 上線執行以前的重要前置準備</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/aa-testing-before-ab-testing/featured.png"/></item><item><title>是不是常態分佈？Q-Q Plot 一看就懂</title><link>https://haosquare.com/normal-distribution-qqplot/</link><pubDate>Fri, 21 May 2021 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/normal-distribution-qqplot/</guid><description>檢驗資料是否為常態分佈的視覺化方法：Q-Q Plot，附上 R 語言程式碼教學如何製圖</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/normal-distribution-qqplot/featured.png"/></item><item><title>A/B Test：「偷看結果」將成為最大的錯誤</title><link>https://haosquare.com/ab-testing-peeking/</link><pubDate>Sat, 22 Aug 2020 00:00:00 +0000</pubDate><author>haosquare.tw@gmail.com (好豪)</author><guid>https://haosquare.com/ab-testing-peeking/</guid><description>在實驗樣本蒐集完整之前「偷看結果」，最嚴重的下場，會讓你手中看似有統計顯著性的決策，20% 以上是錯誤的！</description><media:content xmlns:media="http://search.yahoo.com/mrss/" url="https://haosquare.com/ab-testing-peeking/featured.jpg"/></item></channel></rss>