Timeline for Autocorrelation in linear mixed models (lme)
Current License: CC BY-SA 4.0
15 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Sep 24, 2023 at 16:58 | comment | added | Chris S. | I'm not sure this is appropriate, but could you calculate first difference to see if the autocorrelation is removed? You would lose one observation per whale. | |
| Sep 24, 2023 at 15:07 | answer | added | DrJerryTAO | timeline score: 1 | |
| Feb 21, 2021 at 1:25 | comment | added | Gavin Simpson |
It looks as if you have used the default residuals where you need the normalized residuals to include the effect of the covariance structure; do resid(model, type = "normalized") to access the required residuals
|
|
| Feb 21, 2021 at 0:00 | history | tweeted | twitter.com/StackStats/status/1363277321696915471 | ||
| Jul 10, 2020 at 8:29 | comment | added | Catarina Toscano | Meanwhile I tried using corARMA and it solved the problem of autocorrelation! | |
| Jul 8, 2020 at 14:08 | comment | added | Catarina Toscano | Regarding the other variables, my goal was to see what varibles varyed according to dive type (including the influence of diel because previous research stated that the whales dive behaviour is differs between night and day). Is it a wrong approach to model each metric separately? | |
| Jul 8, 2020 at 14:06 | comment | added | Catarina Toscano | Thanks, I will try to include the interaction. Regarding the actual date/day I haven't included that variable. My data is quite spread in terms of date/year (1 or 2 whales per year where tagged) so I don't think I can use season or year as factor (not representative enough of each season/year). | |
| Jul 8, 2020 at 12:39 | comment | added | Robert Long |
Another question :) Only dive_type and diel_1 are included in your model, but you have several other variables that vary with each dive. Some of the autocorrelation (and whale & residual variance) may be accounted byt those those variables. Is there any reason why you haven't included them ?
|
|
| Jul 8, 2020 at 11:23 | comment | added | Robert Long |
OK thanks, I understand what diel is now, but isn't there another time variable for the actual date/day ? What was the syntax of the model which included time for which you got the error ? Were you just specifying corAR1(form = ~ diel | whale) ? That definitely won't work. Instead you need something like corAR1(form = ~ diel:day | whale)
|
|
| Jul 8, 2020 at 10:42 | comment | added | user289381 | Agree with Robert. If you have repeated measures at different time points, you need to model the interaction between time and type of dive. From the information available is not clear if you have enough data to model time as a continuous variable or it's better discretising it. | |
| Jul 8, 2020 at 10:22 | comment | added | Catarina Toscano | Hello Robert, I added an explanation of diel as asked. Thanks! | |
| Jul 8, 2020 at 10:20 | history | edited | Catarina Toscano | CC BY-SA 4.0 |
added 490 characters in body
|
| Jul 8, 2020 at 4:48 | comment | added | Robert Long | What is the time/lag variable here ? I know in your other question you mentioned that you could not include time in the corAR1 function because of repeated diel values - please can you edit the question and explain this a bit more ? If the measurements are repeated at different times of the day, you might need to specify the interection between day and diel, but I'm not sure I am understanding the design properly yet. | |
| Jul 7, 2020 at 13:52 | history | edited | Catarina Toscano | CC BY-SA 4.0 |
added 68 characters in body
|
| Jul 7, 2020 at 13:43 | history | asked | Catarina Toscano | CC BY-SA 4.0 |