Timeline for Best practices in the selection of distance metric and clustering methods for gene expression data
Current License: CC BY-SA 4.0
7 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Apr 8, 2020 at 20:48 | comment | added | Atakan | Thanks so much, makes sense | |
| Apr 8, 2020 at 14:18 | vote | accept | Atakan | ||
| Apr 8, 2020 at 10:36 | comment | added | Karolis Koncevičius | @Atakan likely all the combinations of distances and linkage methods should be reasoned about individually, considering what happens in each separate case. Ward's method tries to minimise within cluster sum os squares at each step, which makes sense when used on euclidean distances. But correlation distance has a monotonic relationship with euclidean distances, if the values are centered and scaled (as described HERE). Hence you can center, scale, and use euclidean distance to mimic the correlation-based approache. | |
| Apr 8, 2020 at 10:21 | comment | added | Atakan | This has gotta be one of the most understandable answers to questions regarding clustering. Thank you very much. I'd appreciate it you can comment on the validity of using certain methods and distances together. In various places I read that Ward's method requires squared euclidean distance. Algorithms won't complain and can still perform clustering. Is there a combination I should never use? | |
| Apr 7, 2020 at 23:02 | history | edited | Karolis Koncevičius | CC BY-SA 4.0 |
edited body
|
| Apr 7, 2020 at 22:49 | history | edited | Karolis Koncevičius | CC BY-SA 4.0 |
deleted 3 characters in body
|
| Apr 7, 2020 at 22:42 | history | answered | Karolis Koncevičius | CC BY-SA 4.0 |