From 520328538ec265223b05e659e570b7dbbf9ae498 Mon Sep 17 00:00:00 2001 From: s444501 Date: Wed, 18 May 2022 06:51:04 +0200 Subject: [PATCH] test 3 --- bootstrap-t.ipynb | 68 ++++++++++++++++++++++++----------------------- 1 file changed, 35 insertions(+), 33 deletions(-) diff --git a/bootstrap-t.ipynb b/bootstrap-t.ipynb index e6eec77..0f694a4 100644 --- a/bootstrap-t.ipynb +++ b/bootstrap-t.ipynb @@ -23,6 +23,19 @@ "Wszystkie rodzaje testów są testami parametrycznymi, a co za tym idzie nasze mierzone zmienne ilościowe powinny mieć rozkład normalny." ] }, + { + "cell_type": "markdown", + "source": [ + "## Testowanie hipotez metodą bootstrap\n", + "\n", + "**Bootstrap** – metoda szacowania (estymacji) wyników poprzez wielokrotne losowanie ze zwracaniem z próby. Polega ona na utworzeniu nowego rozkładu wyników, na podstawie posiadanych danych, poprzez wielokrotne losowanie wartości z posiadanej próby. Metoda ze zwracaniem polega na tym, że po wylosowaniu danej wartości, “wraca” ona z powrotem do zbioru.\n", + "\n", + "Metoda bootstrapowa znajduje zastosowanie w sytuacji, w której nie znamy rozkładu z populacji z której pochodzi próbka lub w przypadku rozkładów małych lub asymetrycznych. W takim wypadku, dzięki tej metodzie, wyniki testów parametrycznych i analiz opartych o modele liniowe są bardziej precyzyjne. Zazwyczaj losuje się wiele próbek, np. 2000 czy 5000." + ], + "metadata": { + "collapsed": false + } + }, { "cell_type": "markdown", "source": [ @@ -34,7 +47,7 @@ }, { "cell_type": "code", - "execution_count": 546, + "execution_count": 582, "metadata": { "pycharm": { "name": "#%%\n" @@ -51,7 +64,7 @@ }, { "cell_type": "code", - "execution_count": 547, + "execution_count": 583, "metadata": {}, "outputs": [], "source": [ @@ -60,7 +73,7 @@ }, { "cell_type": "code", - "execution_count": 548, + "execution_count": 584, "metadata": {}, "outputs": [], "source": [ @@ -71,7 +84,7 @@ }, { "cell_type": "code", - "execution_count": 549, + "execution_count": 585, "metadata": {}, "outputs": [], "source": [ @@ -92,7 +105,7 @@ }, { "cell_type": "code", - "execution_count": 550, + "execution_count": 586, "metadata": { "pycharm": { "name": "#%%\n" @@ -114,7 +127,7 @@ }, { "cell_type": "code", - "execution_count": 551, + "execution_count": 587, "metadata": {}, "outputs": [], "source": [ @@ -132,7 +145,7 @@ }, { "cell_type": "code", - "execution_count": 552, + "execution_count": 588, "metadata": {}, "outputs": [], "source": [ @@ -150,7 +163,7 @@ }, { "cell_type": "code", - "execution_count": 553, + "execution_count": 589, "metadata": {}, "outputs": [], "source": [ @@ -178,7 +191,7 @@ }, { "cell_type": "code", - "execution_count": 554, + "execution_count": 590, "metadata": {}, "outputs": [], "source": [ @@ -217,7 +230,7 @@ }, { "cell_type": "code", - "execution_count": 555, + "execution_count": 591, "metadata": {}, "outputs": [ { @@ -268,20 +281,9 @@ "\n" ] }, - { - "cell_type": "markdown", - "metadata": {}, - "source": [ - "## Testowanie hipotez metodą bootstrap\n", - "\n", - "**Bootstrap** – metoda szacowania (estymacji) wyników poprzez wielokrotne losowanie ze zwracaniem z próby. Polega ona na utworzeniu nowego rozkładu wyników, na podstawie posiadanych danych, poprzez wielokrotne losowanie wartości z posiadanej próby. Metoda ze zwracaniem polega na tym, że po wylosowaniu danej wartości, “wraca” ona z powrotem do zbioru.\n", - "\n", - "Metoda bootstrapowa znajduje zastosowanie w sytuacji, w której nie znamy rozkładu z populacji z której pochodzi próbka lub w przypadku rozkładów małych lub asymetrycznych. W takim wypadku, dzięki tej metodzie, wyniki testów parametrycznych i analiz opartych o modele liniowe są bardziej precyzyjne. Zazwyczaj losuje się wiele próbek, np. 2000 czy 5000." - ] - }, { "cell_type": "code", - "execution_count": 556, + "execution_count": 592, "metadata": { "pycharm": { "name": "#%%\n" @@ -309,7 +311,7 @@ }, { "cell_type": "code", - "execution_count": 557, + "execution_count": 593, "metadata": { "collapsed": false, "pycharm": { @@ -344,7 +346,7 @@ }, { "cell_type": "code", - "execution_count": 558, + "execution_count": 594, "metadata": { "collapsed": false, "pycharm": { @@ -366,7 +368,7 @@ }, { "cell_type": "code", - "execution_count": 559, + "execution_count": 595, "metadata": { "collapsed": false, "pycharm": { @@ -388,7 +390,7 @@ }, { "cell_type": "code", - "execution_count": 560, + "execution_count": 596, "metadata": { "collapsed": false, "pycharm": { @@ -421,7 +423,7 @@ }, { "cell_type": "code", - "execution_count": 561, + "execution_count": 597, "outputs": [], "source": [ "dataset = pd.read_csv('experiment_data.csv')\n", @@ -461,7 +463,7 @@ }, { "cell_type": "code", - "execution_count": 561, + "execution_count": 597, "outputs": [], "source": [], "metadata": { @@ -505,7 +507,7 @@ }, { "cell_type": "code", - "execution_count": 561, + "execution_count": 597, "outputs": [], "source": [], "metadata": { @@ -562,7 +564,7 @@ }, { "cell_type": "code", - "execution_count": 562, + "execution_count": 598, "outputs": [ { "name": "stdout", @@ -603,7 +605,7 @@ }, { "cell_type": "code", - "execution_count": 563, + "execution_count": 599, "outputs": [ { "name": "stdout", @@ -617,7 +619,7 @@ "p: 1.0\n", "Wartość statystyki testowej z próby: [7.89079918]\n", "Wartości statystyk z prób boostrapowych:\n", - "[-2.17000034], [-0.74957325], [-1.53238091], [-2.4791557], [1.17261618], ... (i 95 pozostałych)\n", + "[-1.2615733], [-0.73536146], [0.56657145], [-0.63034854], [0.27066658], ... (i 95 pozostałych)\n", "\n", "\n" ] @@ -625,7 +627,7 @@ { "data": { "text/plain": "
", - "image/png": "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\n" + "image/png": "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\n" }, "metadata": { "needs_background": "light"