Class "unmarkedFit"
unmarkedFit-class.RdContains fitted model information which can be manipulated or extracted using the methods described below.
Slots
fitType:Object of class
"character"call:Object of class
"call"formula:Object of class
"formula"data:Object of class
"unmarkedFrame"sitesRemoved:Object of class
"numeric"estimates:Object of class
"unmarkedEstimateList"AIC:Object of class
"numeric"opt:Object of class
"list"containing results fromoptimnegLogLike:Object of class
"numeric"nllFun:Object of class
"function"knownOcc:unmarkedFitOccu only: sites known to be occupied
K:unmarkedFitPCount only: upper bound used in integration
mixture:unmarkedFitPCount only: Mixing distribution
keyfun:unmarkedFitDS only: detection function used by distsamp
unitsOut:unmarkedFitDS only: density units
Methods
- [
signature(x = "unmarkedFit", i = "ANY", j = "ANY", drop = "ANY"): extract one of names(obj), eg 'state' or 'det'- backTransform
signature(obj = "unmarkedFit"): back-transform parameters to original scale when no covariate effects are modeled- coef
signature(object = "unmarkedFit"): returns parameter estimates. type can be one of names(obj), eg 'state' or 'det'. If altNames=TRUE estimate names are more specific.- confint
signature(object = "unmarkedFit"): Returns confidence intervals. Must specify type and method (either "normal" or "profile")- fitted
signature(object = "unmarkedFit"): returns expected values of Y- getData
signature(object = "unmarkedFit"): extracts data- getP
signature(object = "unmarkedFit"): calculates and extracts expected detection probabilities- getFP
signature(object = "unmarkedFit"): calculates and extracts expected false positive detection probabilities- getB
signature(object = "unmarkedFit"): calculates and extracts expected probabilities a true positive detection was classified as certain- hessian
signature(object = "unmarkedFit"): Returns hessian matrix- linearComb
signature(obj = "unmarkedFit", coefficients = "matrixOrVector"): Returns estimate and SE on original scale when covariates are present- mle
signature(object = "unmarkedFit"): Same as coef(fit)?- names
signature(x = "unmarkedFit"): Names of parameter levels- nllFun
signature(object = "unmarkedFit"): returns negative log-likelihood used to estimate parameters- parboot
signature(object = "unmarkedFit"): Parametric bootstrapping method to assess goodness-of-fit- plot
signature(x = "unmarkedFit", y = "missing"): Plots expected vs. observed values- predict
signature(object = "unmarkedFit"): Returns predictions and standard errors for original data or for covariates in a new data.frame- profile
signature(fitted = "unmarkedFit"): used by confint method='profile'- residuals
signature(object = "unmarkedFit"): returns residuals- sampleSize
signature(object = "unmarkedFit"): returns number of sites in sample- SE
signature(obj = "unmarkedFit"): returns standard errors- show
signature(object = "unmarkedFit"): concise results- summary
signature(object = "unmarkedFit"): results with more details- update
signature(object = "unmarkedFit"): refit model with changes to one or more arguments- vcov
signature(object = "unmarkedFit"): returns variance-covariance matrix- smoothed
signature(object="unmarkedFitColExt"): Returns the smoothed trajectory from a colonization-extinction model fit. Takes additional logical argument mean which specifies whether or not to return the average over sites.- projected
signature(object="unmarkedFitColExt"): Returns the projected trajectory from a colonization-extinction model fit. Takes additional logical argument mean which specifies whether or not to return the average over sites.- logLik
signature(object="unmarkedFit"): Returns the log-likelihood.- LRT
signature(m1="unmarkedFit", m2="unmarkedFit"): Returns the chi-squared statistic, degrees-of-freedom, and p-value from a Likelihood Ratio Test.
Note
This is a superclass with child classes for each fit type
Examples
showClass("unmarkedFit")
#> Class "unmarkedFit" [package "unmarked"]
#>
#> Slots:
#>
#> Name: fitType call formula
#> Class: character call formula
#>
#> Name: data sitesRemoved estimates
#> Class: unmarkedFrame numeric unmarkedEstimateList
#>
#> Name: AIC opt negLogLike
#> Class: numeric list numeric
#>
#> Name: nllFun bootstrapSamples covMatBS
#> Class: function optionalList optionalMatrix
#>
#> Name: TMB
#> Class: optionalList
#>
#> Known Subclasses:
#> Class "unmarkedFitDS", directly
#> Class "unmarkedFitPCount", directly
#> Class "unmarkedFitDailMadsen", directly
#> Class "unmarkedFitOccu", directly
#> Class "unmarkedFitOccuPEN", directly
#> Class "unmarkedFitOccuPEN_CV", directly
#> Class "unmarkedFitOccuFP", directly
#> Class "unmarkedFitOccuMulti", directly
#> Class "unmarkedFitOccuMS", directly
#> Class "unmarkedFitOccuTTD", directly
#> Class "unmarkedFitNmixTTD", directly
#> Class "unmarkedFitMPois", directly
#> Class "unmarkedFitOccuRN", directly
#> Class "unmarkedFitMNmix", directly
#> Class "unmarkedFitColExt", directly
#> Class "unmarkedFitGMM", directly
#> Class "unmarkedFitIDS", directly
#> Class "unmarkedFitGOccu", directly
#> Class "unmarkedFitOccuCOP", directly
#> Class "unmarkedFitPCO", by class "unmarkedFitDailMadsen", distance 2
#> Class "unmarkedFitMMO", by class "unmarkedFitDailMadsen", distance 2
#> Class "unmarkedFitDSO", by class "unmarkedFitDailMadsen", distance 2
#> Class "unmarkedFitGDS", by class "unmarkedFitGMM", distance 2
#> Class "unmarkedFitGPC", by class "unmarkedFitGMM", distance 2
#> Class "unmarkedFitGDR", by class "unmarkedFitGMM", distance 3
# Format removal data for multinomPois
data(ovendata)
ovenFrame <- unmarkedFrameMPois(y = ovendata.list$data,
siteCovs = as.data.frame(scale(ovendata.list$covariates[,-1])),
type = "removal")
# Fit a couple of models
(fm1 <- multinomPois(~ 1 ~ ufc + trba, ovenFrame))
#>
#> Call:
#> multinomPois(formula = ~1 ~ ufc + trba, data = ovenFrame)
#>
#> Abundance:
#> Estimate SE z P(>|z|)
#> (Intercept) 0.102 0.119 0.864 0.388
#> ufc 0.100 0.126 0.794 0.427
#> trba -0.171 0.135 -1.262 0.207
#>
#> Detection:
#> Estimate SE z P(>|z|)
#> 0.288 0.233 1.24 0.217
#>
#> AIC: 326.1387
summary(fm1)
#>
#> Call:
#> multinomPois(formula = ~1 ~ ufc + trba, data = ovenFrame)
#>
#> Abundance (log-scale):
#> Estimate SE z P(>|z|)
#> (Intercept) 0.102 0.119 0.864 0.388
#> ufc 0.100 0.126 0.794 0.427
#> trba -0.171 0.135 -1.262 0.207
#>
#> Detection (logit-scale):
#> Estimate SE z P(>|z|)
#> 0.288 0.233 1.24 0.217
#>
#> AIC: 326.1387
#> Number of sites: 70
#> optim convergence code: 0
#> optim iterations: 30
#> Bootstrap iterations: 0
#>
# Apply a bunch of methods to the fitted model
# Look at the different parameter types
names(fm1)
#> [1] "state" "det"
fm1['state']
#> Abundance:
#> Estimate SE z P(>|z|)
#> (Intercept) 0.102 0.119 0.864 0.388
#> ufc 0.100 0.126 0.794 0.427
#> trba -0.171 0.135 -1.262 0.207
fm1['det']
#> Detection:
#> Estimate SE z P(>|z|)
#> 0.288 0.233 1.24 0.217
# Coefficients from abundance part of the model
coef(fm1, type='state')
#> lambda(Int) lambda(ufc) lambda(trba)
#> 0.1023681 0.1002939 -0.1708609
# Variance-covariance matrix
vcov(fm1, type='state')
#> lambda(Int) lambda(ufc) lambda(trba)
#> lambda(Int) 0.014052472 -0.001228179 0.002364962
#> lambda(ufc) -0.001228179 0.015941971 0.008057614
#> lambda(trba) 0.002364962 0.008057614 0.018318119
# Confidence intervals using profiled likelihood
confint(fm1, type='state', method='profile')
#> Profiling parameter 1 of 3 ... done.
#> Profiling parameter 2 of 3 ... done.
#> Profiling parameter 3 of 3 ... done.
#> 0.025 0.975
#> lambda(Int) -0.1390786 0.32676614
#> lambda(ufc) -0.1477724 0.34811770
#> lambda(trba) -0.4368444 0.09469605
# Expected values
fitted(fm1)
#> [,1] [,2] [,3] [,4]
#> [1,] 0.8960021 0.3840242 0.16459179 0.07054362
#> [2,] 0.7870851 0.3373426 0.14458420 0.06196842
#> [3,] 0.8518174 0.3650868 0.15647527 0.06706490
#> [4,] 0.9330928 0.3999212 0.17140521 0.07346383
#> [5,] 0.6193105 0.2654349 0.11376473 0.04875927
#> [6,] 0.6709329 0.2875601 0.12324754 0.05282358
#> [7,] 0.4425821 0.1896896 0.08130047 0.03484517
#> [8,] 0.6406227 0.2745692 0.11767968 0.05043721
#> [9,] 0.5722107 0.2452480 0.10511269 0.04505103
#> [10,] 0.7758598 0.3325315 0.14252217 0.06108464
#> [11,] 0.6696637 0.2870162 0.12301439 0.05272365
#> [12,] 0.7332182 0.3142555 0.13468909 0.05772740
#> [ reached getOption("max.print") -- omitted 58 rows ]
# Original data
getData(fm1)
#> Data frame representation of unmarkedFrame object.
#> y.1 y.2 y.3 y.4 ufc trba
#> 1 0 0 0 0 0.84177975 -1.5395877
#> 2 1 0 0 0 0.03022367 -1.2574150
#> 3 0 0 0 0 2.10015927 -0.5049547
#> 4 0 0 0 0 1.08597395 -1.6336452
#> 5 0 0 0 0 1.00492226 0.7177934
#> 6 0 0 0 0 0.84177975 0.1534482
#> 7 0 0 0 0 -0.90290994 1.5643114
#> 8 0 0 0 0 -0.74080655 -0.5049547
#> [ reached 'max' / getOption("max.print") -- omitted 62 rows ]
# Detection probabilities
getP(fm1)
#> [,1] [,2] [,3] [,4]
#> [1,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [2,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [3,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [4,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [5,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [6,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [7,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [8,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [9,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [10,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [11,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [12,] 0.5714025 0.2449017 0.1049642 0.0449874
#> [ reached getOption("max.print") -- omitted 58 rows ]
# log-likelihood
logLik(fm1)
#> [1] -159.0694
# Back-transform detection probability to original scale
# backTransform only works on models with no covariates or
# in conjunction with linearComb (next example)
backTransform(fm1, type ='det')
#> Backtransformed linear combination(s) of Detection estimate(s)
#>
#> Estimate SE LinComb (Intercept)
#> 0.571 0.057 0.288 1
#>
#> Transformation: logistic
# Predicted abundance at specified covariate values
(lc <- linearComb(fm1, c(Int = 1, ufc = 0, trba = 0), type='state'))
#> Linear combination(s) of Abundance estimate(s)
#>
#> Estimate SE (Intercept) ufc trba
#> 0.102 0.119 1 0 0
#>
backTransform(lc)
#> Backtransformed linear combination(s) of Abundance estimate(s)
#>
#> Estimate SE LinComb (Intercept) ufc trba
#> 1.11 0.131 0.102 1 0 0
#>
#> Transformation: exp
# Assess goodness-of-fit
parboot(fm1)
#>
| | 0 % ~calculating
|+++++ | 10% ~00s
|++++++++++ | 20% ~00s
|+++++++++++++++ | 30% ~00s
|++++++++++++++++++++ | 40% ~00s
|+++++++++++++++++++++++++ | 50% ~00s
|++++++++++++++++++++++++++++++ | 60% ~00s
|+++++++++++++++++++++++++++++++++++ | 70% ~00s
|++++++++++++++++++++++++++++++++++++++++ | 80% ~00s
|+++++++++++++++++++++++++++++++++++++++++++++ | 90% ~00s
|++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=00s
#>
#> Call: parboot(object = fm1)
#>
#> Parametric Bootstrap Statistics:
#> t0 mean(t0 - t_B) StdDev(t0 - t_B) Pr(t_B > t0)
#> SSE 72.2 -6.09 11.9 0.636
#>
#> t_B quantiles:
#> 0% 2.5% 25% 50% 75% 97.5% 100%
#> [1,] 56 57 72 83 85 92 93
#>
#> t0 = Original statistic computed from data
#> t_B = Vector of bootstrap samples
#>
plot(fm1)
# Predict abundance at specified covariate values.
newdat <- data.frame(ufc = 0, trba = seq(-1, 1, length=10))
predict(fm1, type='state', newdata=newdat)
#> Predicted SE lower upper
#> 1 1.3142011 0.2184923 0.9487362 1.820448
#> 2 1.2652375 0.1853259 0.9494939 1.685978
#> 3 1.2180982 0.1591867 0.9428516 1.573698
#> 4 1.1727152 0.1412680 0.9260934 1.485013
#> 5 1.1290230 0.1324045 0.8971796 1.420778
#> 6 1.0869586 0.1322528 0.8563382 1.379688
#> 7 1.0464615 0.1390828 0.8064770 1.357858
#> 8 1.0074732 0.1505623 0.7516666 1.350336
#> 9 0.9699374 0.1646399 0.6954379 1.352786
#> 10 0.9338002 0.1798639 0.6401766 1.362097
# Number of sites in the sample
sampleSize(fm1)
#> [1] 70
# Fit a new model without covariates
(fmNull <- update(fm1, formula = ~1 ~1))
#>
#> Call:
#> multinomPois(formula = ~1 ~ 1, data = ovenFrame)
#>
#> Abundance:
#> Estimate SE z P(>|z|)
#> 0.13 0.115 1.12 0.262
#>
#> Detection:
#> Estimate SE z P(>|z|)
#> 0.288 0.233 1.24 0.217
#>
#> AIC: 326.2803
# Likelihood ratio test
LRT(fm1, fmNull)
#> Chisq DF Pr(>Chisq)
#> 1 4.141604 2 0.1260846