Why Meta-Analysis Needs a Taxonomy | HubMeta Tutorial #16 — Transcript

Full transcript

1. 0:15Welcome back. Now that we have extracted
2. 0:19the data that we want to include from
3. 0:21our PDFs, it is time for us to do a
4. 0:25little bit of more uh preparation on
5. 0:28that data and then get ready for
6. 0:30analysis. That's what is happening in
7. 0:33the fourth stage of our pipeline which
8. 0:36is push. The first thing and the most
9. 0:39important part is taxonomy. This part is
10. 0:44of course mostly relevant for
11. 0:48uh quantitative systematic reviews aka
12. 0:52metaanalysis projects and uh let me
13. 0:56explain why we need a taxonomy and why
14. 1:00we need to organize that. Imagine in
15. 1:02this study of
16. 1:05companies and firms that have
17. 1:07internationalized and we are looking at
18. 1:09their level of um their international
19. 1:13sales and the effect of that on their
20. 1:15performance. One of the things that many
21. 1:18many studies have measured is a simple
22. 1:22things such as firm size like usually
23. 1:25this one is measured as the number of
24. 1:28people that work in that company. But if
25. 1:30I type in uh firm size here like this is
26. 1:34the list of all of these measurements.
27. 1:36You can see that people have used many
28. 1:40many different ways of measuring the
29. 1:43same concept. So some of them are using
30. 1:47logarithmic measures, some of them are
31. 1:50using you know sales, some of them using
32. 1:53assets, number of people, many many
33. 1:56different ways
34. 1:59to include them in a metaanalytic
35. 2:04uh analysis.
36. 2:06You have to make decisions
37. 2:09as to how do we group these measurements
38. 2:14under the same parent construct. So in
39. 2:17this example, my parent construct would
40. 2:20become firm size which is a construct
41. 2:25that measures how big or how small this
42. 2:28company is. The question of how this
43. 2:32firm size construct is measured is
44. 2:36answered by different people in many
45. 2:38different ways. You might have hundreds
46. 2:41of different ways of measuring firm
47. 2:43size.
48. 2:45So what happens in meta analysis is that
49. 2:49the grouping strategy, the grouping
50. 2:52decisions that you have to make to put
51. 2:55this measurements into a final
52. 2:59meaningfully coherent set under the same
53. 3:03parent is a very difficult one. And to
54. 3:08do that efficiently with thousands of
55. 3:11included articles is a big project of
56. 3:14its own. That's where our taxonomy page
57. 3:19comes into play. Let's review our
58. 3:21taxonomy page in general and then we
59. 3:24will go into more details. First of all,
60. 3:28you have three views in our taxonomy
61. 3:30page. One of them is a table view where
62. 3:33you can see the different you know
63. 3:36measurements, the numbers when they were
64. 3:38added. You can add by uh to see all the
65. 3:43ones that have already been assigned to
66. 3:45a parent for example, all the ones that
67. 3:49are not assigned for example or orphaned
68. 3:52right now and all the parents that do
69. 3:56not have you know something attached to
70. 3:58them. Or you can select by constructs or
71. 4:05measurements or both like you know
72. 4:07different filters are available to you
73. 4:10and you also have this tree view which
74. 4:13we have created. This tree view allows
75. 4:15you to see like how the different uh
76. 4:19constructs are attached to their uh
77. 4:23child measurements. Like in this case,
78. 4:25marketing intensity as our con construct
79. 4:28is measured through multiple different
80. 4:31ways. You can play around with this to
81. 4:33see if any of these views is more or
82. 4:36less relevant to you. But the explorer
83. 4:40view which we show by default is the one
84. 4:42that I think is the most useful one.
85. 4:45Here you can open and close your
86. 4:47construct and this will clearly show you
87. 4:51what measurements have been categorized
88. 4:53under that. So in this project because I
89. 4:55have already run the uh assignment
90. 4:58things. Uh if I open my firm size, I
91. 5:01will see like 28 different measurements
92. 5:05across my papers that have have been
93. 5:09used to define uh to capture uh firm
94. 5:13size. Any taxonomy node including a
95. 5:17construct or measurement you you click
96. 5:20on inside hub meta will open a right
97. 5:23pane which is node inspector. It shows
98. 5:26the name of the node, the the exact
99. 5:28definition, the type it is, uh the
100. 5:33parent node and you know the
101. 5:35relationship with the parent that they
102. 5:38have. It could be a positive or negative
103. 5:40uh relationship.
104. 5:43Here you can export all of these into a
105. 5:47CSV file. So all of your measurements
106. 5:49you can just take them out of Hubmeta
107. 5:50and do something. And here you can also
108. 5:53import u from an existing project or you
109. 5:57know you will do your own categorization
110. 5:59manually and then bring it back to hub
111. 6:01meta. So we have allowed for that you
112. 6:04can manually add nodes define a
113. 6:06construct or measurement here
114. 6:09and you know use the fields. uh but the
115. 6:14part that is very important inside
116. 6:16taxonomy is this taxonomy wizard which I
117. 6:21will explain in the next video. The
118. 6:24taxonomy wizard is where we will
119. 6:28use the power of AI to uh attach our
120. 6:32measurements to uh meaningful constructs
121. 6:37into meaningful groups which will then
122. 6:39allow us to analyze this project.