Optimal RM in ENMeval

[Jane Anderson]

Hi All,

I'm using ENMeval to create CEMs of the native ranges of a few invasive species to then project into introduced habitat. I was able to get the ENMeval code based on the vignette to work without any problems, however I'd like to then run the optimal model (based on AICc) through Maxent so I can project into novel habitat. I can't tell how to discern what RM was used for the optimal model. If anyone can provide guidance on this I would really appreciate it!

[Jamie M. Kass]

If “e” is your ENMeval object, then e@results will give you the results table that tells you all the evalution stats (i.e. AUC, AIC, omission rates, etc.). It also tells you the feature class and reg multiplier combo for each model. Once you know which model you’d like to go forward with, e@models[[n]], where n is the row number of the model in the results table, will give you the model object you can use to project to other areas with the predict() function in dismo.

Jamie Kass
PhD Candidate
City College of NY

[Jane Anderson]

Thank you so much! I really appreciate the info. 

--
You received this message because you are subscribed to the Google Groups "Maxent" group.
To unsubscribe from this group and stop receiving emails from it, send an email to maxent+unsubscribe@googlegroups.com.
To post to this group, send email to max...@googlegroups.com.
Visit this group at https://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/d/optout.

--

C. Jane Anderson, Ph.D.

www.ecojane.org

[Jane Anderson]

Hi Jamie,

I'm working on trying to apply the optimal model to novel habitat. Following your directions, I used e@results to compare the models, and I am now trying to use  e@models[[n]] to apply the optimal model to the new habitat rasters. I initially included 8 bioclim variables, the optimal model includes 5. When I use  e@models[[4]] (with 4 being the optimal model), it returns all 8 parameters. When I call in the lambdas for this model, it does show the appropriate lambas for the AICc=0 model (three of the variables have lambda = 0). My understanding of this is that the model still includes all 8 original parameters, but places lambda=0 for the three parameters not included in the optimal model. Is that correct? Output is below in case it is helpful. Thank you so much for your guidance!

> Mfeval1@results 

  settings features  rm full.AUC  Mean.AUC     Var.AUC Mean.AUC.DIFF Var.AUC.DIFF Mean.OR10 Var.OR10 Mean.ORmin

1    L_0.5        L 0.5   0.8120 0.8225286 0.001630585   0.006598828 0.0005194964        NA       NA 0.01470588

2      L_1        L 1.0   0.8116 0.8235455 0.001445713   0.005438728 0.0003528934        NA       NA 0.01470588

3    L_1.5        L 1.5   0.8112 0.8245692 0.001246707   0.004035665 0.0001943028        NA       NA 0.01470588

4      L_2        L 2.0   0.8110 0.8249912 0.001080381   0.003002312 0.0001075377        NA       NA 0.01470588

     Var.ORmin     AICc delta.AICc      w.AIC nparam

1 0.0008650519 4089.818   6.474702 0.02810316      8

2 0.0008650519 4089.887   6.543478 0.02715318      8

3 0.0008650519 4085.622   2.278467 0.22906364      6

4 0.0008650519 4083.344   0.000000 0.71568003      5

> Mfeval1@models[[4]]

class    : MaxEnt 

variables: bio_12 bio_13 bio_15 bio_18 bio_19 bio_3 bio_4 bio_5 

output html file no longer exists

> Mfeval1@models[[4]]@lambdas

 [1] "bio_12, 0.0, 812.842102050781, 5635.8701171875"                

 [2] "bio_13, 0.0, 119.75, 1303.13000488281"                         

 [3] "bio_15, 2.074535745678483, 9.0, 117.160003662109"              

 [4] "bio_18, -3.564961303151841, 139.839996337891, 1526.68005371094"

 [5] "bio_19, -0.042996764383110266, 5.0, 2151.0"                    

 [6] "bio_3, 0.0, 46.0416717529297, 93.7200012207031"                

 [7] "bio_4, -4.642562384517404, 146.0, 2724.8798828125"             

 [8] "bio_5, -0.4453853617290973, 166.199996948242, 386.880004882812"

 [9] "linearPredictorNormalizer, 0.41074341844497464"                

[10] "densityNormalizer, 705.5873942373165"                          

[11] "numBackgroundPoints, 9534"                                     

[12] "entropy, 8.479327617339674"                                    

> 

On Fri, Apr 20, 2018 at 5:13 AM, Jamie M. Kass <ndimhy...@gmail.com> wrote:

--
You received this message because you are subscribed to the Google Groups "Maxent" group.
To unsubscribe from this group and stop receiving emails from it, send an email to maxent+unsubscribe@googlegroups.com.
To post to this group, send email to max...@googlegroups.com.
Visit this group at https://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/d/optout.

[Jamie M. Kass]

Yes, the console output from printing the ENMeval object will tell you how many variables went into the model, not how many had coefficients greater than 0. Also keep in mind that with feature classes more complex than Linear, features of variables many make in into the model (e.g. some Hinge features), while the linear component, or the one without textual qualifiers like tick marks, etc. denoting more complex features, may be 0. Your models were all Linear only, so we wouldn’t see this here.

Jamie Kass

PhD Candidate

City College of NY

To unsubscribe from this group and stop receiving emails from it, send an email to maxent+un...@googlegroups.com.

To post to this group, send email to max...@googlegroups.com.
Visit this group at https://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/d/optout.

--

C. Jane Anderson, Ph.D.

www.ecojane.org

--
You received this message because you are subscribed to a topic in the Google Groups "Maxent" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/maxent/8akE0Xv87Gg/unsubscribe.
To unsubscribe from this group and all its topics, send an email to maxent+un...@googlegroups.com.

To post to this group, send email to max...@googlegroups.com.
Visit this group at https://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/d/optout.

--

-----------------------------------------------------------
Jamie M. Kass
PhD Candidate, Dept. of Biology
City College of New York, CUNY Graduate Center