Script qualité habitat Lézard ocellé
Mathis Damestoy
2026-08-20
R Markdown
library(readr)
Maille_lézard_habitats <- read_csv(
"Maille_lézard_habitats.csv",
na = "NA"
)## Rows: 389 Columns: 14
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## dbl (14): id, left, top, right, bottom, row_index, col_index, Présence, Terr...
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.Maille_lézard_habitats## # A tibble: 389 × 14
## id left top right bottom row_index col_index Présence Terriers
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 227 -155020. 5443002. -154920. 5.44e6 72 2 0 0
## 2 228 -155020. 5442902. -154920. 5.44e6 73 2 0 0
## 3 229 -155020. 5442802. -154920. 5.44e6 74 2 0 0
## 4 230 -155020. 5442702. -154920. 5.44e6 75 2 0 0
## 5 296 -154920. 5443802. -154820. 5.44e6 64 3 0 0
## 6 297 -154920. 5443702. -154820. 5.44e6 65 3 0 0
## 7 298 -154920. 5443602. -154820. 5.44e6 66 3 0 0
## 8 299 -154920. 5443502. -154820. 5.44e6 67 3 0 0
## 9 300 -154920. 5443402. -154820. 5.44e6 68 3 0 0
## 10 301 -154920. 5443302. -154820. 5.44e6 69 3 0 0
## # ℹ 379 more rows
## # ℹ 5 more variables: Refuge_art <dbl>, Bosquet <dbl>, `% dune gri` <dbl>,
## # distance <dbl>, distance_2 <dbl>data <- Maille_lézard_habitats
library(readr)
table(data$Présence)##
## 0 1
## 329 60#Distance aux terriers
boxplot(distance ~ Présence,
data = data)
#Distance refuges artificiels
boxplot(distance_2 ~ Présence,
data = data)
#Distance dune grise
boxplot(`% dune gri` ~ Présence,
data = data)
#Test
table(data$Présence,
data$Terriers)##
## 0 1
## 0 313 16
## 1 44 16chisq.test(data$Présence,
data$Terriers)## Warning in chisq.test(data$Présence, data$Terriers): L’approximation du Chi-2
## est peut-être incorrecte##
## Pearson's Chi-squared test with Yates' continuity correction
##
## data: data$Présence and data$Terriers
## X-squared = 29.131, df = 1, p-value = 6.763e-08Les mailles comportant au moins un terrier présentent environ 4 fois plus de chances d’être occupées par le Lézard ocellé que les mailles sans terrier. La présence de terriers est fortement associée à la présence du Lézard ocellé (χ² = 29,13 ; ddl = 1 ; p < 0,001).
library(epitools)
oddsratio(table(data$Présence,
data$Terriers))## Warning in chisq.test(xx, correct = correction): L’approximation du Chi-2 est
## peut-être incorrecte## $data
##
## 0 1 Total
## 0 313 16 329
## 1 44 16 60
## Total 357 32 389
##
## $measure
## odds ratio with 95% C.I.
## estimate lower upper
## 0 1.000000 NA NA
## 1 7.052621 3.262809 15.30129
##
## $p.value
## two-sided
## midp.exact fisher.exact chi.square
## 0 NA NA NA
## 1 1.558015e-06 1.37964e-06 1.578457e-08
##
## $correction
## [1] FALSE
##
## attr(,"method")
## [1] "median-unbiased estimate & mid-p exact CI"Les chances d’observer le Lézard ocellé sont environ 7 à 8 fois plus élevées dans les mailles comportant un terrier.
chisq.test(data$Présence,
data$Refuge_art)##
## Pearson's Chi-squared test with Yates' continuity correction
##
## data: data$Présence and data$Refuge_art
## X-squared = 24.583, df = 1, p-value = 7.118e-07Significatif
oddsratio(table(data$Présence,
data$Refuge_art))## $data
##
## 0 1 Total
## 0 302 27 329
## 1 41 19 60
## Total 343 46 389
##
## $measure
## odds ratio with 95% C.I.
## estimate lower upper
## 0 1.000000 NA NA
## 1 5.157144 2.603401 10.11699
##
## $p.value
## two-sided
## midp.exact fisher.exact chi.square
## 0 NA NA NA
## 1 5.305534e-06 4.501672e-06 2.273345e-07
##
## $correction
## [1] FALSE
##
## attr(,"method")
## [1] "median-unbiased estimate & mid-p exact CI"chance 5 fois plus élevé dans les refuges art
chisq.test(data$Présence,
data$Bosquet) ##
## Pearson's Chi-squared test with Yates' continuity correction
##
## data: data$Présence and data$Bosquet
## X-squared = 3.4982, df = 1, p-value = 0.06144Pas significatif
wilcox.test(distance ~ Présence,
data = data)##
## Wilcoxon rank sum test with continuity correction
##
## data: distance by Présence
## W = 14074, p-value = 1.543e-07
## alternative hypothesis: true location shift is not equal to 0Significatif
wilcox.test(distance_2 ~ Présence,
data = data)##
## Wilcoxon rank sum test with continuity correction
##
## data: distance_2 by Présence
## W = 14780, p-value = 8.814e-10
## alternative hypothesis: true location shift is not equal to 0Hautement significatif la distance au refuge
wilcox.test(data$Présence, data$`% dune gri`)##
## Wilcoxon rank sum test with continuity correction
##
## data: data$Présence and data$`% dune gri`
## W = 66230, p-value = 1.841e-05
## alternative hypothesis: true location shift is not equal to 0#Sélectionner uniquement les variables explicatives
vars <- data[,c("Terriers",
"Refuge_art",
"% dune gri",
"distance",
"distance_2")]
cor(vars)## Terriers Refuge_art % dune gri distance distance_2
## Terriers 1.000000000 0.006256684 0.48000069 -0.32760009 -0.17625463
## Refuge_art 0.006256684 1.000000000 0.06593862 0.17935837 -0.38499838
## % dune gri 0.480000694 0.065938621 1.00000000 -0.32484180 -0.24843079
## distance -0.327600093 0.179358373 -0.32484180 1.00000000 0.08129464
## distance_2 -0.176254627 -0.384998385 -0.24843079 0.08129464 1.00000000#premier GLM
mod1 <- glm(
Présence ~ Terriers +
Refuge_art +
`% dune gri` +
distance +
distance_2,
family = binomial,
data = data
)
summary(mod1)##
## Call:
## glm(formula = Présence ~ Terriers + Refuge_art + `% dune gri` +
## distance + distance_2, family = binomial, data = data)
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.3475019 0.3394940 -3.969 7.21e-05 ***
## Terriers 1.0390727 0.4804886 2.163 0.030577 *
## Refuge_art 1.4642441 0.4417001 3.315 0.000916 ***
## `% dune gri` 0.0127312 0.0077120 1.651 0.098772 .
## distance -0.0014157 0.0006035 -2.346 0.018996 *
## distance_2 -0.0008391 0.0003683 -2.278 0.022727 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 334.54 on 388 degrees of freedom
## Residual deviance: 266.01 on 383 degrees of freedom
## AIC: 278.01
##
## Number of Fisher Scoring iterations: 6#Sélection de modèles
library(MuMIn)
options(na.action = "na.fail")
mods <- dredge(mod1)## Fixed term is "(Intercept)"mods## Global model call: glm(formula = Présence ~ Terriers + Refuge_art + `% dune gri` +
## distance + distance_2, family = binomial, data = data)
## ---
## Model selection table
## (Int) `%dungri` dst dst_2 Rfg_art Trr df logLik AICc
## 32 -1.3480 0.01273 -0.001416 -0.0008391 1.4640 1.0390 6 -133.005 278.2
## 31 -1.1470 -0.001629 -0.0009340 1.4650 1.3380 5 -134.349 278.9
## 16 -1.1540 0.01896 -0.001703 -0.0008722 1.4210 5 -135.306 280.8
## 28 -1.8930 0.01673 -0.001593 2.0190 1.1210 5 -136.109 282.4
## 30 -1.7800 0.01736 -0.0009029 1.1680 1.3190 5 -136.197 282.6
## 27 -1.7010 -0.001898 2.0890 1.5320 4 -138.386 284.9
## 12 -1.6920 0.02371 -0.001952 1.9960 4 -138.739 285.6
## 15 -0.7177 -0.002223 -0.0010440 1.3760 4 -138.823 285.8
## 29 -1.5880 -0.0010410 1.1150 1.8060 4 -138.857 285.8
## 24 -0.9368 0.01270 -0.000974 -0.0013780 0.9431 5 -138.668 287.5
## 26 -2.4400 0.02199 1.7540 1.4650 4 -139.892 287.9
## 14 -1.6220 0.02659 -0.0009810 1.0460 4 -140.072 288.2
## 23 -0.7328 -0.001197 -0.0014800 1.2320 4 -140.118 288.3
## 22 -1.3040 0.01610 -0.0013540 1.1600 4 -140.232 288.6
## 8 -0.7708 0.01798 -0.001264 -0.0013920 4 -140.689 289.5
## 21 -1.1450 -0.0014720 1.6160 3 -142.642 291.3
## 6 -1.2130 0.02420 -0.0013800 3 -143.411 292.9
## 25 -2.3140 1.7950 2.1240 3 -144.087 294.2
## 7 -0.3592 -0.001804 -0.0015530 3 -144.140 294.3
## 11 -1.2500 -0.002699 2.0420 3 -144.168 294.4
## 10 -2.3240 0.03299 1.6780 3 -144.643 295.3
## 13 -1.1740 -0.0012670 0.8559 3 -147.973 302.0
## 5 -0.8622 -0.0015950 2 -150.458 304.9
## 20 -1.6670 0.01952 -0.001216 1.0090 4 -148.608 305.3
## 18 -2.1050 0.02329 1.2880 3 -150.411 306.9
## 4 -1.4600 0.02478 -0.001659 3 -150.931 307.9
## 19 -1.4090 -0.001636 1.4440 3 -152.092 310.2
## 2 -2.0200 0.03237 2 -154.422 312.9
## 17 -1.9620 1.9620 2 -155.466 315.0
## 9 -1.9970 1.6450 2 -156.722 317.5
## 3 -0.9462 -0.002641 2 -157.563 319.2
## 1 -1.7020 1 -167.269 336.5
## delta weight
## 32 0.00 0.416
## 31 0.63 0.304
## 16 2.54 0.117
## 28 4.15 0.052
## 30 4.32 0.048
## 27 6.65 0.015
## 12 7.35 0.011
## 15 7.52 0.010
## 29 7.59 0.009
## 24 9.26 0.004
## 26 9.66 0.003
## 14 10.02 0.003
## 23 10.11 0.003
## 22 10.34 0.002
## 8 11.25 0.001
## 21 13.12 0.001
## 6 14.65 0.000
## 25 16.01 0.000
## 7 16.11 0.000
## 11 16.17 0.000
## 10 17.12 0.000
## 13 23.78 0.000
## 5 26.72 0.000
## 20 27.09 0.000
## 18 28.66 0.000
## 4 29.69 0.000
## 19 32.02 0.000
## 2 34.65 0.000
## 17 36.73 0.000
## 9 39.24 0.000
## 3 40.93 0.000
## 1 58.32 0.000
## Models ranked by AICc(x)best <- get.models(mods,1)[[1]]
summary(best)##
## Call:
## glm(formula = Présence ~ `% dune gri` + distance + distance_2 +
## Refuge_art + Terriers + 1, family = binomial, data = data)
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.3475019 0.3394940 -3.969 7.21e-05 ***
## `% dune gri` 0.0127312 0.0077120 1.651 0.098772 .
## distance -0.0014157 0.0006035 -2.346 0.018996 *
## distance_2 -0.0008391 0.0003683 -2.278 0.022727 *
## Refuge_art 1.4642441 0.4417001 3.315 0.000916 ***
## Terriers 1.0390727 0.4804886 2.163 0.030577 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 334.54 on 388 degrees of freedom
## Residual deviance: 266.01 on 383 degrees of freedom
## AIC: 278.01
##
## Number of Fisher Scoring iterations: 6#Vérification de la performance prédictive
library(pROC)## Warning: le package 'pROC' a été compilé avec la version R 4.5.3## Type 'citation("pROC")' for a citation.##
## Attachement du package : 'pROC'## Les objets suivants sont masqués depuis 'package:stats':
##
## cov, smooth, varpred <- predict(best,
type = "response")
roc_obj <- roc(data$Présence,
pred)## Setting levels: control = 0, case = 1## Setting direction: controls < casesauc(roc_obj)## Area under the curve: 0.8013plot(roc_obj)
exp(coef(best))## (Intercept) `% dune gri` distance distance_2 Refuge_art Terriers
## 0.2598887 1.0128126 0.9985853 0.9991613 4.3242734 2.8265948#Validation croisé
library(caret)## Warning: le package 'caret' a été compilé avec la version R 4.5.3## Le chargement a nécessité le package : ggplot2## Le chargement a nécessité le package : latticectrl <- trainControl(
method = "cv",
number = 10,
classProbs = TRUE,
summaryFunction = twoClassSummary)#Vérification de la colinéarité
cor(data[,c("Terriers","distance")])## Terriers distance
## Terriers 1.0000000 -0.3276001
## distance -0.3276001 1.0000000library(car)## Le chargement a nécessité le package : carDatavif(best)## `% dune gri` distance distance_2 Refuge_art Terriers
## 1.376568 1.320044 1.306099 1.375633 1.358826#Production de la carte de favorabilité des habitats
data$favorabilite <- predict(best,
type = "response")
summary(data$favorabilite)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.02353 0.03719 0.10552 0.15424 0.15843 0.84559hist(data$favorabilite,
breaks = 20)
#Enregistre données
write.csv(data,
"favorabilite_ocelle.csv",
row.names = FALSE)
write.csv(
data[,c("id","favorabilite")],
"favorabilite.csv",
row.names = FALSE
)
getwd()## [1] "C:/Users/mathi/Downloads"