Ramon Meffert
commited on
Commit
•
e7b8106
1
Parent(s):
fffa7ce
Add timings analysis
Browse files- results/f1_analysis.ipynb +1 -1
- results/timings_analysis.ipynb +635 -0
results/f1_analysis.ipynb
CHANGED
@@ -440,7 +440,7 @@
|
|
440 |
"cell_type": "markdown",
|
441 |
"metadata": {},
|
442 |
"source": [
|
443 |
-
"From these results, we can see that both the retriever and the reader have a significant effect on the F1 score ($F = 58.63$ and $F = 16.23$ respectively, $p < 0.0001$ for both). However, there is also an interaction between the retriever and reader ($F = 43.53$, $p < 0.0001$). The post-hoc analysis of contrasts shows that ElasticSearch performs better than FAISS ($p < 0.0001$) and Longformer performs better than DPR ($p
|
444 |
]
|
445 |
}
|
446 |
],
|
|
|
440 |
"cell_type": "markdown",
|
441 |
"metadata": {},
|
442 |
"source": [
|
443 |
+
"From these results, we can see that both the retriever and the reader have a significant effect on the F1 score ($F = 58.63$ and $F = 16.23$ respectively, $p < 0.0001$ for both). However, there is also an interaction between the retriever and reader ($F = 43.53$, $p < 0.0001$). The post-hoc analysis of contrasts shows that ElasticSearch performs better than FAISS ($p < 0.0001$) and Longformer performs better than DPR ($p = 0.0001$)."
|
444 |
]
|
445 |
}
|
446 |
],
|
results/timings_analysis.ipynb
ADDED
@@ -0,0 +1,635 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cells": [
|
3 |
+
{
|
4 |
+
"cell_type": "markdown",
|
5 |
+
"metadata": {},
|
6 |
+
"source": [
|
7 |
+
"# Timings"
|
8 |
+
]
|
9 |
+
},
|
10 |
+
{
|
11 |
+
"cell_type": "code",
|
12 |
+
"execution_count": 1,
|
13 |
+
"metadata": {
|
14 |
+
"vscode": {
|
15 |
+
"languageId": "r"
|
16 |
+
}
|
17 |
+
},
|
18 |
+
"outputs": [
|
19 |
+
{
|
20 |
+
"name": "stderr",
|
21 |
+
"output_type": "stream",
|
22 |
+
"text": [
|
23 |
+
"Loading required package: ggplot2\n",
|
24 |
+
"\n",
|
25 |
+
"-- \u001b[1mAttaching packages\u001b[22m --------------------------------------- tidyverse 1.3.1 --\n",
|
26 |
+
"\n",
|
27 |
+
"\u001b[32mv\u001b[39m \u001b[34mtibble \u001b[39m 3.1.5 \u001b[32mv\u001b[39m \u001b[34mdplyr \u001b[39m 1.0.7\n",
|
28 |
+
"\u001b[32mv\u001b[39m \u001b[34mtidyr \u001b[39m 1.1.4 \u001b[32mv\u001b[39m \u001b[34mstringr\u001b[39m 1.4.0\n",
|
29 |
+
"\u001b[32mv\u001b[39m \u001b[34mpurrr \u001b[39m 0.3.4 \u001b[32mv\u001b[39m \u001b[34mforcats\u001b[39m 0.5.1\n",
|
30 |
+
"\n",
|
31 |
+
"-- \u001b[1mConflicts\u001b[22m ------------------------------------------ tidyverse_conflicts() --\n",
|
32 |
+
"\u001b[31mx\u001b[39m \u001b[34mdplyr\u001b[39m::\u001b[32mfilter()\u001b[39m masks \u001b[34mstats\u001b[39m::filter()\n",
|
33 |
+
"\u001b[31mx\u001b[39m \u001b[34mdplyr\u001b[39m::\u001b[32mlag()\u001b[39m masks \u001b[34mstats\u001b[39m::lag()\n",
|
34 |
+
"\n",
|
35 |
+
"Loading required package: mvtnorm\n",
|
36 |
+
"\n",
|
37 |
+
"Loading required package: survival\n",
|
38 |
+
"\n",
|
39 |
+
"Loading required package: TH.data\n",
|
40 |
+
"\n",
|
41 |
+
"Loading required package: MASS\n",
|
42 |
+
"\n",
|
43 |
+
"\n",
|
44 |
+
"Attaching package: 'MASS'\n",
|
45 |
+
"\n",
|
46 |
+
"\n",
|
47 |
+
"The following object is masked from 'package:dplyr':\n",
|
48 |
+
"\n",
|
49 |
+
" select\n",
|
50 |
+
"\n",
|
51 |
+
"\n",
|
52 |
+
"\n",
|
53 |
+
"Attaching package: 'TH.data'\n",
|
54 |
+
"\n",
|
55 |
+
"\n",
|
56 |
+
"The following object is masked from 'package:MASS':\n",
|
57 |
+
"\n",
|
58 |
+
" geyser\n",
|
59 |
+
"\n",
|
60 |
+
"\n",
|
61 |
+
"Loading required package: carData\n",
|
62 |
+
"\n",
|
63 |
+
"\n",
|
64 |
+
"Attaching package: 'car'\n",
|
65 |
+
"\n",
|
66 |
+
"\n",
|
67 |
+
"The following object is masked from 'package:dplyr':\n",
|
68 |
+
"\n",
|
69 |
+
" recode\n",
|
70 |
+
"\n",
|
71 |
+
"\n",
|
72 |
+
"The following object is masked from 'package:purrr':\n",
|
73 |
+
"\n",
|
74 |
+
" some\n",
|
75 |
+
"\n",
|
76 |
+
"\n",
|
77 |
+
"\n",
|
78 |
+
"Attaching package: 'rstatix'\n",
|
79 |
+
"\n",
|
80 |
+
"\n",
|
81 |
+
"The following object is masked from 'package:MASS':\n",
|
82 |
+
"\n",
|
83 |
+
" select\n",
|
84 |
+
"\n",
|
85 |
+
"\n",
|
86 |
+
"The following object is masked from 'package:stats':\n",
|
87 |
+
"\n",
|
88 |
+
" filter\n",
|
89 |
+
"\n",
|
90 |
+
"\n"
|
91 |
+
]
|
92 |
+
}
|
93 |
+
],
|
94 |
+
"source": [
|
95 |
+
"library(\"ggpubr\")\n",
|
96 |
+
"library(readr)\n",
|
97 |
+
"library(ggplot2)\n",
|
98 |
+
"library(tidyverse)\n",
|
99 |
+
"library(ARTool)\n",
|
100 |
+
"library(emmeans)\n",
|
101 |
+
"library(multcomp)\n",
|
102 |
+
"library(car)\n",
|
103 |
+
"library(rstatix)"
|
104 |
+
]
|
105 |
+
},
|
106 |
+
{
|
107 |
+
"cell_type": "code",
|
108 |
+
"execution_count": 2,
|
109 |
+
"metadata": {
|
110 |
+
"vscode": {
|
111 |
+
"languageId": "r"
|
112 |
+
}
|
113 |
+
},
|
114 |
+
"outputs": [
|
115 |
+
{
|
116 |
+
"name": "stderr",
|
117 |
+
"output_type": "stream",
|
118 |
+
"text": [
|
119 |
+
"New names:\n",
|
120 |
+
"* `` -> ...1\n",
|
121 |
+
"\n",
|
122 |
+
"\u001b[1mRows: \u001b[22m\u001b[34m59\u001b[39m \u001b[1mColumns: \u001b[22m\u001b[34m9\u001b[39m\n",
|
123 |
+
"\u001b[36m--\u001b[39m \u001b[1mColumn specification\u001b[22m \u001b[36m--------------------------------------------------------\u001b[39m\n",
|
124 |
+
"\u001b[1mDelimiter:\u001b[22m \",\"\n",
|
125 |
+
"\u001b[32mdbl\u001b[39m (9): ...1, faiss_dpr.retrieve, faiss_dpr.read, faiss_longformer.retrieve...\n",
|
126 |
+
"\n",
|
127 |
+
"\u001b[36mi\u001b[39m Use `spec()` to retrieve the full column specification for this data.\n",
|
128 |
+
"\u001b[36mi\u001b[39m Specify the column types or set `show_col_types = FALSE` to quiet this message.\n"
|
129 |
+
]
|
130 |
+
},
|
131 |
+
{
|
132 |
+
"data": {
|
133 |
+
"text/html": [
|
134 |
+
"<table class=\"dataframe\">\n",
|
135 |
+
"<caption>A tibble: 6 × 5</caption>\n",
|
136 |
+
"<thead>\n",
|
137 |
+
"\t<tr><th scope=col>question</th><th scope=col>retriever</th><th scope=col>reader</th><th scope=col>method</th><th scope=col>time</th></tr>\n",
|
138 |
+
"\t<tr><th scope=col><dbl></th><th scope=col><fct></th><th scope=col><fct></th><th scope=col><fct></th><th scope=col><dbl></th></tr>\n",
|
139 |
+
"</thead>\n",
|
140 |
+
"<tbody>\n",
|
141 |
+
"\t<tr><td>0</td><td>faiss</td><td>dpr </td><td>retrieve</td><td>0.30384302</td></tr>\n",
|
142 |
+
"\t<tr><td>0</td><td>faiss</td><td>dpr </td><td>read </td><td>4.56640005</td></tr>\n",
|
143 |
+
"\t<tr><td>0</td><td>faiss</td><td>longformer</td><td>retrieve</td><td>0.92279482</td></tr>\n",
|
144 |
+
"\t<tr><td>0</td><td>faiss</td><td>longformer</td><td>read </td><td>5.76836824</td></tr>\n",
|
145 |
+
"\t<tr><td>0</td><td>es </td><td>dpr </td><td>retrieve</td><td>0.01930094</td></tr>\n",
|
146 |
+
"\t<tr><td>0</td><td>es </td><td>dpr </td><td>read </td><td>2.74536490</td></tr>\n",
|
147 |
+
"</tbody>\n",
|
148 |
+
"</table>\n"
|
149 |
+
],
|
150 |
+
"text/latex": [
|
151 |
+
"A tibble: 6 × 5\n",
|
152 |
+
"\\begin{tabular}{lllll}\n",
|
153 |
+
" question & retriever & reader & method & time\\\\\n",
|
154 |
+
" <dbl> & <fct> & <fct> & <fct> & <dbl>\\\\\n",
|
155 |
+
"\\hline\n",
|
156 |
+
"\t 0 & faiss & dpr & retrieve & 0.30384302\\\\\n",
|
157 |
+
"\t 0 & faiss & dpr & read & 4.56640005\\\\\n",
|
158 |
+
"\t 0 & faiss & longformer & retrieve & 0.92279482\\\\\n",
|
159 |
+
"\t 0 & faiss & longformer & read & 5.76836824\\\\\n",
|
160 |
+
"\t 0 & es & dpr & retrieve & 0.01930094\\\\\n",
|
161 |
+
"\t 0 & es & dpr & read & 2.74536490\\\\\n",
|
162 |
+
"\\end{tabular}\n"
|
163 |
+
],
|
164 |
+
"text/markdown": [
|
165 |
+
"\n",
|
166 |
+
"A tibble: 6 × 5\n",
|
167 |
+
"\n",
|
168 |
+
"| question <dbl> | retriever <fct> | reader <fct> | method <fct> | time <dbl> |\n",
|
169 |
+
"|---|---|---|---|---|\n",
|
170 |
+
"| 0 | faiss | dpr | retrieve | 0.30384302 |\n",
|
171 |
+
"| 0 | faiss | dpr | read | 4.56640005 |\n",
|
172 |
+
"| 0 | faiss | longformer | retrieve | 0.92279482 |\n",
|
173 |
+
"| 0 | faiss | longformer | read | 5.76836824 |\n",
|
174 |
+
"| 0 | es | dpr | retrieve | 0.01930094 |\n",
|
175 |
+
"| 0 | es | dpr | read | 2.74536490 |\n",
|
176 |
+
"\n"
|
177 |
+
],
|
178 |
+
"text/plain": [
|
179 |
+
" question retriever reader method time \n",
|
180 |
+
"1 0 faiss dpr retrieve 0.30384302\n",
|
181 |
+
"2 0 faiss dpr read 4.56640005\n",
|
182 |
+
"3 0 faiss longformer retrieve 0.92279482\n",
|
183 |
+
"4 0 faiss longformer read 5.76836824\n",
|
184 |
+
"5 0 es dpr retrieve 0.01930094\n",
|
185 |
+
"6 0 es dpr read 2.74536490"
|
186 |
+
]
|
187 |
+
},
|
188 |
+
"metadata": {},
|
189 |
+
"output_type": "display_data"
|
190 |
+
}
|
191 |
+
],
|
192 |
+
"source": [
|
193 |
+
"timings <- read_csv(\"timings.csv\") %>%\n",
|
194 |
+
" rename(question = `...1`) %>%\n",
|
195 |
+
" pivot_longer(!question, names_to=c(\"retriever\", \"reader\", \"method\"), names_sep=\"[._]\", values_to=\"time\")\n",
|
196 |
+
"\n",
|
197 |
+
"timings$retriever <- as.factor(timings$retriever)\n",
|
198 |
+
"timings$reader <- as.factor(timings$reader)\n",
|
199 |
+
"timings$method <- as.factor(timings$method)\n",
|
200 |
+
"\n",
|
201 |
+
"head(timings)"
|
202 |
+
]
|
203 |
+
},
|
204 |
+
{
|
205 |
+
"cell_type": "code",
|
206 |
+
"execution_count": 3,
|
207 |
+
"metadata": {
|
208 |
+
"vscode": {
|
209 |
+
"languageId": "r"
|
210 |
+
}
|
211 |
+
},
|
212 |
+
"outputs": [],
|
213 |
+
"source": [
|
214 |
+
"timings_read <- filter(timings, method == \"read\") %>%\n",
|
215 |
+
" select(!method)\n",
|
216 |
+
"timings_retrieve <- filter(timings, method == \"retrieve\") %>%\n",
|
217 |
+
" select(!method)"
|
218 |
+
]
|
219 |
+
},
|
220 |
+
{
|
221 |
+
"cell_type": "markdown",
|
222 |
+
"metadata": {},
|
223 |
+
"source": [
|
224 |
+
"To test which tests we can use, we need to check for normality. For this, we use a Shapiro-Wilk test of normality. As you can see in the results below, all $p$-values are lower than 0.001, so we reject the null-hypothesis of normality and now know that none of the f1-scores are normally distributed."
|
225 |
+
]
|
226 |
+
},
|
227 |
+
{
|
228 |
+
"cell_type": "code",
|
229 |
+
"execution_count": 4,
|
230 |
+
"metadata": {
|
231 |
+
"vscode": {
|
232 |
+
"languageId": "r"
|
233 |
+
}
|
234 |
+
},
|
235 |
+
"outputs": [
|
236 |
+
{
|
237 |
+
"data": {
|
238 |
+
"text/html": [
|
239 |
+
"<table class=\"dataframe\">\n",
|
240 |
+
"<caption>A tibble: 2 × 3</caption>\n",
|
241 |
+
"<thead>\n",
|
242 |
+
"\t<tr><th scope=col>retriever</th><th scope=col>sw.stat</th><th scope=col>sw.p</th></tr>\n",
|
243 |
+
"\t<tr><th scope=col><fct></th><th scope=col><dbl></th><th scope=col><dbl></th></tr>\n",
|
244 |
+
"</thead>\n",
|
245 |
+
"<tbody>\n",
|
246 |
+
"\t<tr><td>es </td><td>0.7534261</td><td>1.667341e-18</td></tr>\n",
|
247 |
+
"\t<tr><td>faiss</td><td>0.7585727</td><td>2.563192e-18</td></tr>\n",
|
248 |
+
"</tbody>\n",
|
249 |
+
"</table>\n"
|
250 |
+
],
|
251 |
+
"text/latex": [
|
252 |
+
"A tibble: 2 × 3\n",
|
253 |
+
"\\begin{tabular}{lll}\n",
|
254 |
+
" retriever & sw.stat & sw.p\\\\\n",
|
255 |
+
" <fct> & <dbl> & <dbl>\\\\\n",
|
256 |
+
"\\hline\n",
|
257 |
+
"\t es & 0.7534261 & 1.667341e-18\\\\\n",
|
258 |
+
"\t faiss & 0.7585727 & 2.563192e-18\\\\\n",
|
259 |
+
"\\end{tabular}\n"
|
260 |
+
],
|
261 |
+
"text/markdown": [
|
262 |
+
"\n",
|
263 |
+
"A tibble: 2 × 3\n",
|
264 |
+
"\n",
|
265 |
+
"| retriever <fct> | sw.stat <dbl> | sw.p <dbl> |\n",
|
266 |
+
"|---|---|---|\n",
|
267 |
+
"| es | 0.7534261 | 1.667341e-18 |\n",
|
268 |
+
"| faiss | 0.7585727 | 2.563192e-18 |\n",
|
269 |
+
"\n"
|
270 |
+
],
|
271 |
+
"text/plain": [
|
272 |
+
" retriever sw.stat sw.p \n",
|
273 |
+
"1 es 0.7534261 1.667341e-18\n",
|
274 |
+
"2 faiss 0.7585727 2.563192e-18"
|
275 |
+
]
|
276 |
+
},
|
277 |
+
"metadata": {},
|
278 |
+
"output_type": "display_data"
|
279 |
+
},
|
280 |
+
{
|
281 |
+
"data": {
|
282 |
+
"text/html": [
|
283 |
+
"<table class=\"dataframe\">\n",
|
284 |
+
"<caption>A tibble: 2 × 3</caption>\n",
|
285 |
+
"<thead>\n",
|
286 |
+
"\t<tr><th scope=col>reader</th><th scope=col>sw.stat</th><th scope=col>sw.p</th></tr>\n",
|
287 |
+
"\t<tr><th scope=col><fct></th><th scope=col><dbl></th><th scope=col><dbl></th></tr>\n",
|
288 |
+
"</thead>\n",
|
289 |
+
"<tbody>\n",
|
290 |
+
"\t<tr><td>dpr </td><td>0.7639005</td><td>4.029344e-18</td></tr>\n",
|
291 |
+
"\t<tr><td>longformer</td><td>0.8116362</td><td>3.381683e-16</td></tr>\n",
|
292 |
+
"</tbody>\n",
|
293 |
+
"</table>\n"
|
294 |
+
],
|
295 |
+
"text/latex": [
|
296 |
+
"A tibble: 2 × 3\n",
|
297 |
+
"\\begin{tabular}{lll}\n",
|
298 |
+
" reader & sw.stat & sw.p\\\\\n",
|
299 |
+
" <fct> & <dbl> & <dbl>\\\\\n",
|
300 |
+
"\\hline\n",
|
301 |
+
"\t dpr & 0.7639005 & 4.029344e-18\\\\\n",
|
302 |
+
"\t longformer & 0.8116362 & 3.381683e-16\\\\\n",
|
303 |
+
"\\end{tabular}\n"
|
304 |
+
],
|
305 |
+
"text/markdown": [
|
306 |
+
"\n",
|
307 |
+
"A tibble: 2 × 3\n",
|
308 |
+
"\n",
|
309 |
+
"| reader <fct> | sw.stat <dbl> | sw.p <dbl> |\n",
|
310 |
+
"|---|---|---|\n",
|
311 |
+
"| dpr | 0.7639005 | 4.029344e-18 |\n",
|
312 |
+
"| longformer | 0.8116362 | 3.381683e-16 |\n",
|
313 |
+
"\n"
|
314 |
+
],
|
315 |
+
"text/plain": [
|
316 |
+
" reader sw.stat sw.p \n",
|
317 |
+
"1 dpr 0.7639005 4.029344e-18\n",
|
318 |
+
"2 longformer 0.8116362 3.381683e-16"
|
319 |
+
]
|
320 |
+
},
|
321 |
+
"metadata": {},
|
322 |
+
"output_type": "display_data"
|
323 |
+
},
|
324 |
+
{
|
325 |
+
"data": {
|
326 |
+
"text/html": [
|
327 |
+
"<table class=\"dataframe\">\n",
|
328 |
+
"<caption>A tibble: 2 × 3</caption>\n",
|
329 |
+
"<thead>\n",
|
330 |
+
"\t<tr><th scope=col>method</th><th scope=col>sw.stat</th><th scope=col>sw.p</th></tr>\n",
|
331 |
+
"\t<tr><th scope=col><fct></th><th scope=col><dbl></th><th scope=col><dbl></th></tr>\n",
|
332 |
+
"</thead>\n",
|
333 |
+
"<tbody>\n",
|
334 |
+
"\t<tr><td>read </td><td>0.8838182</td><td>1.779766e-12</td></tr>\n",
|
335 |
+
"\t<tr><td>retrieve</td><td>0.6237773</td><td>1.838892e-22</td></tr>\n",
|
336 |
+
"</tbody>\n",
|
337 |
+
"</table>\n"
|
338 |
+
],
|
339 |
+
"text/latex": [
|
340 |
+
"A tibble: 2 × 3\n",
|
341 |
+
"\\begin{tabular}{lll}\n",
|
342 |
+
" method & sw.stat & sw.p\\\\\n",
|
343 |
+
" <fct> & <dbl> & <dbl>\\\\\n",
|
344 |
+
"\\hline\n",
|
345 |
+
"\t read & 0.8838182 & 1.779766e-12\\\\\n",
|
346 |
+
"\t retrieve & 0.6237773 & 1.838892e-22\\\\\n",
|
347 |
+
"\\end{tabular}\n"
|
348 |
+
],
|
349 |
+
"text/markdown": [
|
350 |
+
"\n",
|
351 |
+
"A tibble: 2 × 3\n",
|
352 |
+
"\n",
|
353 |
+
"| method <fct> | sw.stat <dbl> | sw.p <dbl> |\n",
|
354 |
+
"|---|---|---|\n",
|
355 |
+
"| read | 0.8838182 | 1.779766e-12 |\n",
|
356 |
+
"| retrieve | 0.6237773 | 1.838892e-22 |\n",
|
357 |
+
"\n"
|
358 |
+
],
|
359 |
+
"text/plain": [
|
360 |
+
" method sw.stat sw.p \n",
|
361 |
+
"1 read 0.8838182 1.779766e-12\n",
|
362 |
+
"2 retrieve 0.6237773 1.838892e-22"
|
363 |
+
]
|
364 |
+
},
|
365 |
+
"metadata": {},
|
366 |
+
"output_type": "display_data"
|
367 |
+
}
|
368 |
+
],
|
369 |
+
"source": [
|
370 |
+
"timings %>%\n",
|
371 |
+
" group_by(retriever) %>%\n",
|
372 |
+
" summarise(sw.stat = shapiro.test(time)$statistic,\n",
|
373 |
+
" sw.p = shapiro.test(time)$p)\n",
|
374 |
+
"\n",
|
375 |
+
"timings %>%\n",
|
376 |
+
" group_by(reader) %>%\n",
|
377 |
+
" summarise(sw.stat = shapiro.test(time)$statistic,\n",
|
378 |
+
" sw.p = shapiro.test(time)$p)\n",
|
379 |
+
"\n",
|
380 |
+
"timings %>%\n",
|
381 |
+
" group_by(method) %>%\n",
|
382 |
+
" summarise(sw.stat = shapiro.test(time)$statistic,\n",
|
383 |
+
" sw.p = shapiro.test(time)$p)"
|
384 |
+
]
|
385 |
+
},
|
386 |
+
{
|
387 |
+
"cell_type": "markdown",
|
388 |
+
"metadata": {},
|
389 |
+
"source": [
|
390 |
+
"Since our data is not normally distributed, we cannot use an ANOVA to compare our results. Therefore, we use an aligned-rank test, which is a non-parameteric version of a factorial repeated measures ANOVA."
|
391 |
+
]
|
392 |
+
},
|
393 |
+
{
|
394 |
+
"cell_type": "code",
|
395 |
+
"execution_count": 5,
|
396 |
+
"metadata": {
|
397 |
+
"vscode": {
|
398 |
+
"languageId": "r"
|
399 |
+
}
|
400 |
+
},
|
401 |
+
"outputs": [
|
402 |
+
{
|
403 |
+
"data": {
|
404 |
+
"text/html": [
|
405 |
+
"<table class=\"dataframe\">\n",
|
406 |
+
"<caption>A anova.art: 3 × 7</caption>\n",
|
407 |
+
"<thead>\n",
|
408 |
+
"\t<tr><th></th><th scope=col>Term</th><th scope=col>Df</th><th scope=col>Df.res</th><th scope=col>Sum Sq</th><th scope=col>Sum Sq.res</th><th scope=col>F value</th><th scope=col>Pr(>F)</th></tr>\n",
|
409 |
+
"\t<tr><th></th><th scope=col><chr></th><th scope=col><dbl></th><th scope=col><dbl></th><th scope=col><dbl></th><th scope=col><dbl></th><th scope=col><dbl></th><th scope=col><dbl></th></tr>\n",
|
410 |
+
"</thead>\n",
|
411 |
+
"<tbody>\n",
|
412 |
+
"\t<tr><th scope=row>retriever</th><td>retriever </td><td>1</td><td>232</td><td> 41088.97</td><td>1037631.8</td><td> 9.18692</td><td>2.714084e-03</td></tr>\n",
|
413 |
+
"\t<tr><th scope=row>reader</th><td>reader </td><td>1</td><td>232</td><td>790427.81</td><td> 301414.1</td><td>608.39633</td><td>8.802730e-67</td></tr>\n",
|
414 |
+
"\t<tr><th scope=row>retriever:reader</th><td>retriever:reader</td><td>1</td><td>232</td><td>101903.46</td><td> 983331.4</td><td> 24.04235</td><td>1.771995e-06</td></tr>\n",
|
415 |
+
"</tbody>\n",
|
416 |
+
"</table>\n"
|
417 |
+
],
|
418 |
+
"text/latex": [
|
419 |
+
"A anova.art: 3 × 7\n",
|
420 |
+
"\\begin{tabular}{r|lllllll}\n",
|
421 |
+
" & Term & Df & Df.res & Sum Sq & Sum Sq.res & F value & Pr(>F)\\\\\n",
|
422 |
+
" & <chr> & <dbl> & <dbl> & <dbl> & <dbl> & <dbl> & <dbl>\\\\\n",
|
423 |
+
"\\hline\n",
|
424 |
+
"\tretriever & retriever & 1 & 232 & 41088.97 & 1037631.8 & 9.18692 & 2.714084e-03\\\\\n",
|
425 |
+
"\treader & reader & 1 & 232 & 790427.81 & 301414.1 & 608.39633 & 8.802730e-67\\\\\n",
|
426 |
+
"\tretriever:reader & retriever:reader & 1 & 232 & 101903.46 & 983331.4 & 24.04235 & 1.771995e-06\\\\\n",
|
427 |
+
"\\end{tabular}\n"
|
428 |
+
],
|
429 |
+
"text/markdown": [
|
430 |
+
"\n",
|
431 |
+
"A anova.art: 3 × 7\n",
|
432 |
+
"\n",
|
433 |
+
"| <!--/--> | Term <chr> | Df <dbl> | Df.res <dbl> | Sum Sq <dbl> | Sum Sq.res <dbl> | F value <dbl> | Pr(>F) <dbl> |\n",
|
434 |
+
"|---|---|---|---|---|---|---|---|\n",
|
435 |
+
"| retriever | retriever | 1 | 232 | 41088.97 | 1037631.8 | 9.18692 | 2.714084e-03 |\n",
|
436 |
+
"| reader | reader | 1 | 232 | 790427.81 | 301414.1 | 608.39633 | 8.802730e-67 |\n",
|
437 |
+
"| retriever:reader | retriever:reader | 1 | 232 | 101903.46 | 983331.4 | 24.04235 | 1.771995e-06 |\n",
|
438 |
+
"\n"
|
439 |
+
],
|
440 |
+
"text/plain": [
|
441 |
+
" Term Df Df.res Sum Sq Sum Sq.res F value \n",
|
442 |
+
"retriever retriever 1 232 41088.97 1037631.8 9.18692\n",
|
443 |
+
"reader reader 1 232 790427.81 301414.1 608.39633\n",
|
444 |
+
"retriever:reader retriever:reader 1 232 101903.46 983331.4 24.04235\n",
|
445 |
+
" Pr(>F) \n",
|
446 |
+
"retriever 2.714084e-03\n",
|
447 |
+
"reader 8.802730e-67\n",
|
448 |
+
"retriever:reader 1.771995e-06"
|
449 |
+
]
|
450 |
+
},
|
451 |
+
"metadata": {},
|
452 |
+
"output_type": "display_data"
|
453 |
+
},
|
454 |
+
{
|
455 |
+
"name": "stderr",
|
456 |
+
"output_type": "stream",
|
457 |
+
"text": [
|
458 |
+
"NOTE: Results may be misleading due to involvement in interactions\n",
|
459 |
+
"\n"
|
460 |
+
]
|
461 |
+
},
|
462 |
+
{
|
463 |
+
"data": {
|
464 |
+
"text/plain": [
|
465 |
+
" contrast estimate SE df t.ratio p.value\n",
|
466 |
+
" es - faiss 26.4 8.71 232 3.031 0.0027\n",
|
467 |
+
"\n",
|
468 |
+
"Results are averaged over the levels of: reader "
|
469 |
+
]
|
470 |
+
},
|
471 |
+
"metadata": {},
|
472 |
+
"output_type": "display_data"
|
473 |
+
},
|
474 |
+
{
|
475 |
+
"name": "stderr",
|
476 |
+
"output_type": "stream",
|
477 |
+
"text": [
|
478 |
+
"NOTE: Results may be misleading due to involvement in interactions\n",
|
479 |
+
"\n"
|
480 |
+
]
|
481 |
+
},
|
482 |
+
{
|
483 |
+
"data": {
|
484 |
+
"text/plain": [
|
485 |
+
" contrast estimate SE df t.ratio p.value\n",
|
486 |
+
" dpr - longformer -116 4.69 232 -24.666 <.0001\n",
|
487 |
+
"\n",
|
488 |
+
"Results are averaged over the levels of: retriever "
|
489 |
+
]
|
490 |
+
},
|
491 |
+
"metadata": {},
|
492 |
+
"output_type": "display_data"
|
493 |
+
}
|
494 |
+
],
|
495 |
+
"source": [
|
496 |
+
"model.acc <- art(time ~ retriever * reader, data = timings_read)\n",
|
497 |
+
"anova(model.acc)\n",
|
498 |
+
"art.con(model.acc, ~ retriever)\n",
|
499 |
+
"art.con(model.acc, ~ reader)"
|
500 |
+
]
|
501 |
+
},
|
502 |
+
{
|
503 |
+
"cell_type": "code",
|
504 |
+
"execution_count": 6,
|
505 |
+
"metadata": {
|
506 |
+
"vscode": {
|
507 |
+
"languageId": "r"
|
508 |
+
}
|
509 |
+
},
|
510 |
+
"outputs": [
|
511 |
+
{
|
512 |
+
"data": {
|
513 |
+
"text/html": [
|
514 |
+
"<table class=\"dataframe\">\n",
|
515 |
+
"<caption>A anova.art: 3 × 7</caption>\n",
|
516 |
+
"<thead>\n",
|
517 |
+
"\t<tr><th></th><th scope=col>Term</th><th scope=col>Df</th><th scope=col>Df.res</th><th scope=col>Sum Sq</th><th scope=col>Sum Sq.res</th><th scope=col>F value</th><th scope=col>Pr(>F)</th></tr>\n",
|
518 |
+
"\t<tr><th></th><th scope=col><chr></th><th scope=col><dbl></th><th scope=col><dbl></th><th scope=col><dbl></th><th scope=col><dbl></th><th scope=col><dbl></th><th scope=col><dbl></th></tr>\n",
|
519 |
+
"</thead>\n",
|
520 |
+
"<tbody>\n",
|
521 |
+
"\t<tr><th scope=row>retriever</th><td>retriever </td><td>1</td><td>232</td><td>821516</td><td>240071.9</td><td>793.8944</td><td>7.630526e-77</td></tr>\n",
|
522 |
+
"\t<tr><th scope=row>reader</th><td>reader </td><td>1</td><td>232</td><td>821516</td><td>214935.3</td><td>886.7398</td><td>3.256422e-81</td></tr>\n",
|
523 |
+
"\t<tr><th scope=row>retriever:reader</th><td>retriever:reader</td><td>1</td><td>232</td><td>821516</td><td>215501.6</td><td>884.4096</td><td>4.148583e-81</td></tr>\n",
|
524 |
+
"</tbody>\n",
|
525 |
+
"</table>\n"
|
526 |
+
],
|
527 |
+
"text/latex": [
|
528 |
+
"A anova.art: 3 × 7\n",
|
529 |
+
"\\begin{tabular}{r|lllllll}\n",
|
530 |
+
" & Term & Df & Df.res & Sum Sq & Sum Sq.res & F value & Pr(>F)\\\\\n",
|
531 |
+
" & <chr> & <dbl> & <dbl> & <dbl> & <dbl> & <dbl> & <dbl>\\\\\n",
|
532 |
+
"\\hline\n",
|
533 |
+
"\tretriever & retriever & 1 & 232 & 821516 & 240071.9 & 793.8944 & 7.630526e-77\\\\\n",
|
534 |
+
"\treader & reader & 1 & 232 & 821516 & 214935.3 & 886.7398 & 3.256422e-81\\\\\n",
|
535 |
+
"\tretriever:reader & retriever:reader & 1 & 232 & 821516 & 215501.6 & 884.4096 & 4.148583e-81\\\\\n",
|
536 |
+
"\\end{tabular}\n"
|
537 |
+
],
|
538 |
+
"text/markdown": [
|
539 |
+
"\n",
|
540 |
+
"A anova.art: 3 × 7\n",
|
541 |
+
"\n",
|
542 |
+
"| <!--/--> | Term <chr> | Df <dbl> | Df.res <dbl> | Sum Sq <dbl> | Sum Sq.res <dbl> | F value <dbl> | Pr(>F) <dbl> |\n",
|
543 |
+
"|---|---|---|---|---|---|---|---|\n",
|
544 |
+
"| retriever | retriever | 1 | 232 | 821516 | 240071.9 | 793.8944 | 7.630526e-77 |\n",
|
545 |
+
"| reader | reader | 1 | 232 | 821516 | 214935.3 | 886.7398 | 3.256422e-81 |\n",
|
546 |
+
"| retriever:reader | retriever:reader | 1 | 232 | 821516 | 215501.6 | 884.4096 | 4.148583e-81 |\n",
|
547 |
+
"\n"
|
548 |
+
],
|
549 |
+
"text/plain": [
|
550 |
+
" Term Df Df.res Sum Sq Sum Sq.res F value \n",
|
551 |
+
"retriever retriever 1 232 821516 240071.9 793.8944\n",
|
552 |
+
"reader reader 1 232 821516 214935.3 886.7398\n",
|
553 |
+
"retriever:reader retriever:reader 1 232 821516 215501.6 884.4096\n",
|
554 |
+
" Pr(>F) \n",
|
555 |
+
"retriever 7.630526e-77\n",
|
556 |
+
"reader 3.256422e-81\n",
|
557 |
+
"retriever:reader 4.148583e-81"
|
558 |
+
]
|
559 |
+
},
|
560 |
+
"metadata": {},
|
561 |
+
"output_type": "display_data"
|
562 |
+
},
|
563 |
+
{
|
564 |
+
"name": "stderr",
|
565 |
+
"output_type": "stream",
|
566 |
+
"text": [
|
567 |
+
"NOTE: Results may be misleading due to involvement in interactions\n",
|
568 |
+
"\n"
|
569 |
+
]
|
570 |
+
},
|
571 |
+
{
|
572 |
+
"data": {
|
573 |
+
"text/plain": [
|
574 |
+
" contrast estimate SE df t.ratio p.value\n",
|
575 |
+
" es - faiss -118 4.19 232 -28.176 <.0001\n",
|
576 |
+
"\n",
|
577 |
+
"Results are averaged over the levels of: reader "
|
578 |
+
]
|
579 |
+
},
|
580 |
+
"metadata": {},
|
581 |
+
"output_type": "display_data"
|
582 |
+
},
|
583 |
+
{
|
584 |
+
"name": "stderr",
|
585 |
+
"output_type": "stream",
|
586 |
+
"text": [
|
587 |
+
"NOTE: Results may be misleading due to involvement in interactions\n",
|
588 |
+
"\n"
|
589 |
+
]
|
590 |
+
},
|
591 |
+
{
|
592 |
+
"data": {
|
593 |
+
"text/plain": [
|
594 |
+
" contrast estimate SE df t.ratio p.value\n",
|
595 |
+
" dpr - longformer -118 3.96 232 -29.778 <.0001\n",
|
596 |
+
"\n",
|
597 |
+
"Results are averaged over the levels of: retriever "
|
598 |
+
]
|
599 |
+
},
|
600 |
+
"metadata": {},
|
601 |
+
"output_type": "display_data"
|
602 |
+
}
|
603 |
+
],
|
604 |
+
"source": [
|
605 |
+
"model.acc <- art(time ~ retriever * reader, data = timings_retrieve)\n",
|
606 |
+
"anova(model.acc)\n",
|
607 |
+
"art.con(model.acc, ~ retriever)\n",
|
608 |
+
"art.con(model.acc, ~ reader)"
|
609 |
+
]
|
610 |
+
},
|
611 |
+
{
|
612 |
+
"cell_type": "markdown",
|
613 |
+
"metadata": {},
|
614 |
+
"source": []
|
615 |
+
}
|
616 |
+
],
|
617 |
+
"metadata": {
|
618 |
+
"kernelspec": {
|
619 |
+
"display_name": "R",
|
620 |
+
"language": "R",
|
621 |
+
"name": "ir"
|
622 |
+
},
|
623 |
+
"language_info": {
|
624 |
+
"codemirror_mode": "r",
|
625 |
+
"file_extension": ".r",
|
626 |
+
"mimetype": "text/x-r-source",
|
627 |
+
"name": "R",
|
628 |
+
"pygments_lexer": "r",
|
629 |
+
"version": "4.1.2"
|
630 |
+
},
|
631 |
+
"orig_nbformat": 4
|
632 |
+
},
|
633 |
+
"nbformat": 4,
|
634 |
+
"nbformat_minor": 2
|
635 |
+
}
|