Title: group mean over several axes of array with nans is wrong · Issue #1118 · larray-project/larray · GitHub
Open Graph Title: group mean over several axes of array with nans is wrong · Issue #1118 · larray-project/larray
X Title: group mean over several axes of array with nans is wrong · Issue #1118 · larray-project/larray
Description: What happens is the mean is computed on each axis in turn (mean of mean). When no nans are involved, we get theoretically the same result. In practice, we loose some precision but it was deemed acceptable so far. However, when nans are i...
Open Graph Description: What happens is the mean is computed on each axis in turn (mean of mean). When no nans are involved, we get theoretically the same result. In practice, we loose some precision but it was deemed acc...
X Description: What happens is the mean is computed on each axis in turn (mean of mean). When no nans are involved, we get theoretically the same result. In practice, we loose some precision but it was deemed acc...
Opengraph URL: https://github.com/larray-project/larray/issues/1118
X: @github
Domain: github.com
{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"group mean over several axes of array with nans is wrong","articleBody":"What happens is the mean is computed on each axis in turn (mean of mean). When no nans are involved, we get theoretically the same result. In practice, we loose some precision but it was deemed acceptable so far. However, when nans are involved, the result is significantly wrong.\r\n\r\n```python\r\n>>> arr = Array([[1, 3], [4, nan]], [Axis('a=a0,a1'), Axis('b=b0,b1')])\r\n>>> arr\r\na\\b b0 b1\r\n a0 1.0 3.0\r\n a1 4.0 nan\r\n>>> arr.mean(\"a0,a1 >> a01\", \"b0,b1 >> b01\")\r\n2.75\r\n```\r\nWhile this should be 2.6666... What happens is that it computes:\r\n\r\n```python\r\n>>> (((1 + 4) / 2) + ((3 + 0) / 1)) / 2\r\n2.75\r\n>>> 1/4 + 4/4 + 3/2\r\n2.75\r\n```\r\n\r\nInstead of:\r\n\r\n```python\r\n>>> (1 + 4 + 3) / 3\r\n2.6666666666666665\r\n>>> 1/3 + 4/3 + 3/3\r\n2.6666666666666665\r\n```\r\n\r\nAs a workaround until larray 0.35 is released, I have recommended using:\r\n\r\n```python\r\n>>> # TODO: do not use this function anymore when larray 0.35 will be available\r\n... def nd_mean(array, axes_or_groups):\r\n... \"\"\"\r\n... Computes the mean of array over axes_or_groups.\r\n... \r\n... This function is temporarily necessary because larray versions up to (and including) 0.34.x\r\n... behave badly when computing the means on groups over several dimensions \r\n... when some values are nans. See https://github.com/larray-project/larray/issues/1118\r\n... \"\"\"\r\n... return array.sum(*axes_or_groups) / (~isnan(array)).sum(*axes_or_groups)\r\n>>> nd_mean(arr, (\"a0,a1 >> a01\", \"b0,b1 >> b01\"))\r\n2.6666666666666665\r\n```\r\n","author":{"url":"https://github.com/gdementen","@type":"Person","name":"gdementen"},"datePublished":"2024-10-16T11:19:25Z","url":"https://github.com/larray-project/larray/issues/1118"}
| react-profiling | 0 |
| react-import-map | react,react-dom,react-dom/client,react-dom/profiling,react-is,react/compiler-runtime,react/jsx-dev-runtime,react/jsx-runtime,scheduler@b8488ee2c322 |
| route-pattern | /_view_fragments/issues/show/:user_id/:repository/:id/issue_layout(.:format) |
| route-controller | voltron_issues_fragments |
| route-action | issue_layout |
| fetch-nonce | v2:756e443a-b7fe-2376-5510-fa4d98683c7a |
| current-catalog-service-hash | 81bb79d38c15960b92d99bca9288a9108c7a47b18f2423d0f6438c5b7bcd2114 |
| request-id | C0B8:B1603:2DA22AE:3C3DBF1:6B001F7D |
| html-safe-nonce | 49bcf90bbc3432dd896a9836e670903d69141770db4051d5ef9002cb81e97011 |
| visitor-payload | eyJyZWZlcnJlciI6IiIsInJlcXVlc3RfaWQiOiJDMEI4OkIxNjAzOjJEQTIyQUU6M0MzREJGMTo2QjAwMUY3RCIsInZpc2l0b3JfaWQiOiI4ODIyODY4NDk0MzcxNTI4NTc0IiwicmVnaW9uX2VkZ2UiOiJpYWQiLCJyZWdpb25fcmVuZGVyIjoiaWFkIn0= |
| visitor-hmac | 61b0d790231cf1aa5e2bd7b2b1c5c027315cd6da26acc5aa1b80d6464e6fd3bc |
| hovercard-subject-tag | issue:2591553700 |
| github-keyboard-shortcuts | repository,issues,copilot |
| google-site-verification | Apib7-x98H0j5cPqHWwSMm6dNU4GmODRoqxLiDzdx9I |
| octolytics-url | https://collector.github.com/github/collect |
| analytics-location | / |
| fb:app_id | 1401488693436528 |
| apple-itunes-app | app-id=1477376905, app-argument=https://github.com/_view_fragments/issues/show/larray-project/larray/1118/issue_layout |
| twitter:image | https://opengraph.githubassets.com/d923a737a578f8f0520019e91d41e23ace44e44f2803d51dc2ccf9a1bbd3d42e/larray-project/larray/issues/1118 |
| twitter:card | summary_large_image |
| og:image | https://opengraph.githubassets.com/d923a737a578f8f0520019e91d41e23ace44e44f2803d51dc2ccf9a1bbd3d42e/larray-project/larray/issues/1118 |
| og:image:alt | What happens is the mean is computed on each axis in turn (mean of mean). When no nans are involved, we get theoretically the same result. In practice, we loose some precision but it was deemed acc... |
| og:image:width | 1200 |
| og:image:height | 600 |
| og:site_name | GitHub |
| og:type | object |
| og:author:username | gdementen |
| hostname | github.com |
| expected-hostname | github.com |
| None | 4eb4cba0d69f045374ac15de9666d22b0961e1d555cc3824f36b7734c00c78e0 |
| turbo-cache-control | no-preview |
| go-import | github.com/larray-project/larray git https://github.com/larray-project/larray.git |
| octolytics-dimension-user_id | 26460981 |
| octolytics-dimension-user_login | larray-project |
| octolytics-dimension-repository_id | 16981067 |
| octolytics-dimension-repository_nwo | larray-project/larray |
| octolytics-dimension-repository_public | true |
| octolytics-dimension-repository_is_fork | false |
| octolytics-dimension-repository_network_root_id | 16981067 |
| octolytics-dimension-repository_network_root_nwo | larray-project/larray |
| turbo-body-classes | logged-out env-production page-responsive |
| disable-turbo | false |
| browser-stats-url | https://api.github.com/_private/browser/stats |
| browser-errors-url | https://api.github.com/_private/browser/errors |
| release | 44c658b24abf00491fb7ae5f56d42c9f38cbeab2 |
| ui-target | full |
| theme-color | #1e2327 |
| color-scheme | light dark |
| react-app-name | issues |
Links:
Viewport: width=device-width