Read on Twitter
-Thread-
re:"No statistical support for wolf control and maternal penning as conservation measures for endangered mountain caribou" by Harding et al
https://link.springer.com/article/10.1007/s10531-020-02008-3#article-info
/1
re:"No statistical support for wolf control and maternal penning as conservation measures for endangered mountain caribou" by Harding et al
https://link.springer.com/article/10.1007/s10531-020-02008-3#article-info
/1
We discussed this paper today in our #WiRELab meeting.
Thanks for your hard work: @TobySpribille @Andrew_T_Cook @matbourbonnais @ChrisDarimont @ViktoriaWagner1 - it definitely made our group think deeper about this issue!
I may/probs missed something, but here goes:
/2
Thanks for your hard work: @TobySpribille @Andrew_T_Cook @matbourbonnais @ChrisDarimont @ViktoriaWagner1 - it definitely made our group think deeper about this issue!
I may/probs missed something, but here goes:
/2
Using the author's data, I:
1) dropped "boreal" ecotype [as in the paper, but not in their code]
2) changed 'penning' to 'wolf' [because both happened at the ~same time in KZA].
3) made a dummy factor for wolf control (1) vs (0) pooled all other treatments and controls.
/3
1) dropped "boreal" ecotype [as in the paper, but not in their code]
2) changed 'penning' to 'wolf' [because both happened at the ~same time in KZA].
3) made a dummy factor for wolf control (1) vs (0) pooled all other treatments and controls.
/3
Code below:
-"treatment": as per Harding et al
-"treatment2": treatment, but "penning" changed to "wolf"
-WOLFt: the "wolf" dummy variable (1:yes, 0:no)
-data3 = excluded 'transloc' for same reasons as "boreal" exclusion by Harding et al (i.e. n=1 observation/ factor level)
/4
-"treatment": as per Harding et al
-"treatment2": treatment, but "penning" changed to "wolf"
-WOLFt: the "wolf" dummy variable (1:yes, 0:no)
-data3 = excluded 'transloc' for same reasons as "boreal" exclusion by Harding et al (i.e. n=1 observation/ factor level)
/4
My AIC table below,
Intercept-only (null) model in yellow.
Wolf removal vs. all others is top performing model.
/5
Intercept-only (null) model in yellow.
Wolf removal vs. all others is top performing model.
/5
Summary of top model here:
Wolf removal has a positive (0.224) and significant effect (p<0.01) on caribou pop'n growth vs doing nothing/ moose reduction.
/6
Wolf removal has a positive (0.224) and significant effect (p<0.01) on caribou pop'n growth vs doing nothing/ moose reduction.
/6
I'm not sure how these data support the conclusion that there is NO statistical support (i.e., title of the paper). There is at least SOME support, no? And why was moose reduction (n=4) left off of the title but penning wasn't (n=1)?
['scuse my POTUS hurricane drawing below]
/7
![I'm not sure how these data support the conclusion that there is NO statistical support (i.e., title of the paper). There is at least SOME support, no? And why was moose reduction (n=4) left off of the title but penning wasn't (n=1)?['scuse my POTUS hurricane drawing below]/7](https://pbs.twimg.com/media/EdGlaarWkAEYLbX.png "I'm not sure how these data support the conclusion that there is NO statistical support (i.e., title of the paper). There is at least SOME support, no? And why was moose reduction (n=4) left off of the title but penning wasn't (n=1)?['scuse my POTUS hurricane drawing below]/7")
![I'm not sure how these data support the conclusion that there is NO statistical support (i.e., title of the paper). There is at least SOME support, no? And why was moose reduction (n=4) left off of the title but penning wasn't (n=1)?['scuse my POTUS hurricane drawing below]/7](https://pbs.twimg.com/media/EdGurWDWoAcpYkg.png "I'm not sure how these data support the conclusion that there is NO statistical support (i.e., title of the paper). There is at least SOME support, no? And why was moose reduction (n=4) left off of the title but penning wasn't (n=1)?['scuse my POTUS hurricane drawing below]/7")
['scuse my POTUS hurricane drawing below]
/7
Also, I 'think' the Tukey HSD w/ Bonferroni in Harding et al shows a p<0.05 for wolf vs control. It wasn't discussed in the paper.
Again - this is definitely more than 'no' statistical support - at least as most ecologists would consider it.
/8
Again - this is definitely more than 'no' statistical support - at least as most ecologists would consider it.
/8
[hears whisper...'but ecotype'] -> sure, another variable was top dog. But that doesn't mean there is an absence of _statistical_ support for wolf removal.
Ethics, policy, triage, media/ #scicomm, etc - YES - do that too. This is a really important discussion to have.
/9
Ethics, policy, triage, media/ #scicomm, etc - YES - do that too. This is a really important discussion to have.
/9
Wildlife science (& stats, ethics, funding, policy, etc) needs to be better - the caribou, the wolves, & the people that interact with these spp deserve our very best efforts...
https://albertawilderness.ca/2015-04-00-science-uncertainty-and-ethics-in-the-alberta-wolf-cull-population-ecology-101/
/10
https://albertawilderness.ca/2015-04-00-science-uncertainty-and-ethics-in-the-alberta-wolf-cull-population-ecology-101/
/10
...re: best efforts -its late here, a tweet thread is all I have time for at the moment. Maybe more later - but if so, probs not in the Atlantic :] Either way, would love to hear your thoughts out there.
/end
/end
