Seascape Stats

Data is a data frame off all CTD casts with environmental and acoustic summary statistics.

cast.df = readRDS('../data/processed/cast_df_stats.rds')

The goals of these GLM models are:

Show the relationship between environment and aggregation characteristics
Show the differing relationships between environment and aggregation characteristics between the two seascape classifications.

I am showing model results in tables by using the tab_model function from the library sjPlot. The below is a wrapper function. Feel free to modify it if you think other stats from the GLMs should be displayed.

# wrapper for showing model results
tab_model.wrap <- function(fit){
  tab_model(fit, show.aic = T, show.se=T, show.ci=F)
}

Cluster GLMs

Again not really sure how to to interpret these. But my take was that there is only significant differences between clusters for density and biomass. Is this your take? Matches the plots.

depth ~ temp x cluster

fit <- glm(swarm_mean_depth ~ temp_mean*cluster, data=cast.df, family=Gamma(link='log'))
tab_model.wrap(fit)

	swarm mean depth
Predictors	Estimates	std. Error	p
(Intercept)	2.45	2.02	0.276
temp mean	1.15	0.06	0.005
cluster [B]	0.73	0.85	0.785
temp mean * cluster [B]	1.03	0.07	0.665
Observations	386
R² Nagelkerke	0.061
AIC	3084.635

anova(fit)

## Analysis of Deviance Table
## 
## Model: Gamma, link: log
## 
## Response: swarm_mean_depth
## 
## Terms added sequentially (first to last)
## 
## 
##                   Df Deviance Resid. Df Resid. Dev
## NULL                                385     146.65
## temp_mean          1   4.8412       384     141.80
## cluster            1   2.5797       383     139.22
## temp_mean:cluster  1   0.0740       382     139.15

ggplot(cast.df, aes(temp_mean, swarm_mean_depth, col = cluster)) + 
  geom_point() + 
  geom_smooth(method = "glm", method.args = list(family = Gamma(link = "log")))

## `geom_smooth()` using formula 'y ~ x'

coverage ~ temp x cluster

## use Box-Cox transformation for area 
bc <- MASS::boxcox(area_corrected_area_sum ~ temp_mean*cluster, data = cast.df, plotit = FALSE)
## lambda is 0.2222 for these data - close to 1/3 but residuals are a bit closer to Normal
lambda <- bc$x[which.max(bc$y)] 

## sjPlot::tab_model() doesn't like transformations in model calls, so modify the data.frame
cast.df <- dplyr::mutate(cast.df, 
                         transf_area_sum = ((area_corrected_area_sum^lambda - 1)/lambda))

fit <- glm(transf_area_sum ~ temp_mean*cluster, data=cast.df, family = "gaussian")
tab_model.wrap(fit)

	transf area sum
Predictors	Estimates	std. Error	p
(Intercept)	-1.05	0.73	0.146
temp mean	-0.12	0.04	0.005
cluster [B]	-2.01	1.03	0.050
temp mean * cluster [B]	0.13	0.06	0.038
Observations	386
R²	0.043
AIC	630.281

ggplot(cast.df, aes(temp_mean, transf_area_sum, col = cluster)) + 
  geom_point() + 
  geom_smooth(method = "lm")

## `geom_smooth()` using formula 'y ~ x'

density ~ temp x cluster

fit <- glm(Sv_mean_mean ~ temp_mean*cluster, data=cast.df, family='gaussian')
tab_model.wrap(fit)

	Sv mean mean
Predictors	Estimates	std. Error	p
(Intercept)	-20.72	6.89	0.003
temp mean	-1.93	0.41	<0.001
cluster [B]	-24.88	9.77	0.011
temp mean * cluster [B]	1.33	0.60	0.026
Observations	386
R²	0.093
AIC	2367.825

ggplot(cast.df, aes(temp_mean, Sv_mean_mean, col = cluster)) + geom_point() + geom_smooth(method = "glm")

## `geom_smooth()` using formula 'y ~ x'

Temperature and Salinity

temp ~ salt

fit <- glm(temp_mean ~ salt_mean, data=cast.df, family='gaussian')
tab_model.wrap(fit)

	temp mean
Predictors	Estimates	std. Error	p
(Intercept)	56.58	5.21	<0.001
salt mean	-1.14	0.15	<0.001
Observations	386
R²	0.135
AIC	1032.233

ggplot(cast.df, aes(salt_mean, temp_mean)) + 
  geom_point() + 
  geom_smooth(method = "glm", method.args = list(family='gaussian'))

## `geom_smooth()` using formula 'y ~ x'

temp ~ salt x survey

fit <- glm(temp_mean ~ salt_mean, data = cast.df, family = gaussian)
tab_model.wrap(fit)

	temp mean
Predictors	Estimates	std. Error	p
(Intercept)	56.58	5.21	<0.001
salt mean	-1.14	0.15	<0.001
Observations	386
R²	0.135
AIC	1032.233

summary(fit)

## 
## Call:
## glm(formula = temp_mean ~ salt_mean, family = gaussian, data = cast.df)
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -2.54627  -0.46314  -0.02568   0.47946   3.04236  
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  56.5820     5.2137  10.853  < 2e-16 ***
## salt_mean    -1.1427     0.1477  -7.735 9.22e-14 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.8402562)
## 
##     Null deviance: 372.93  on 385  degrees of freedom
## Residual deviance: 322.66  on 384  degrees of freedom
## AIC: 1032.2
## 
## Number of Fisher Scoring iterations: 2

ggplot(cast.df, aes(salt_mean, temp_mean, col = survey)) + 
  geom_point() + 
  geom_smooth(method = "glm")

## `geom_smooth()` using formula 'y ~ x'

Crossshore Gradient

temp ~ crossShoreDistance

df.split <- split(cast.df, cast.df$survey)
for (i in 1:length(df.split)){
  message('\n', names(df.split)[i])
  fit <- glm(temp_mean ~ crossShoreDistance, data=df.split[[i]])
  #print(tab_model.wrap(fit))
  print(summary(fit))
}

## 
## 2015_S1

## 
## Call:
## glm(formula = temp_mean ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -1.07877  -0.58608  -0.06332   0.35528   1.58023  
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        17.08930    0.18105  94.388  < 2e-16 ***
## crossShoreDistance  0.13760    0.02265   6.076 9.88e-07 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.396428)
## 
##     Null deviance: 26.925  on 32  degrees of freedom
## Residual deviance: 12.289  on 31  degrees of freedom
## AIC: 67.053
## 
## Number of Fisher Scoring iterations: 2

## 
## 2016_S1

## 
## Call:
## glm(formula = temp_mean ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -1.08708  -0.32856   0.00993   0.28729   0.93718  
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        16.42650    0.07707 213.133  < 2e-16 ***
## crossShoreDistance  0.09207    0.01052   8.755 7.44e-13 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.1710508)
## 
##     Null deviance: 25.086  on 71  degrees of freedom
## Residual deviance: 11.974  on 70  degrees of freedom
## AIC: 81.162
## 
## Number of Fisher Scoring iterations: 2

## 
## 2016_S2

## 
## Call:
## glm(formula = temp_mean ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -0.93236  -0.28289  -0.07889   0.41275   0.81125  
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        15.799954   0.091310 173.037   <2e-16 ***
## crossShoreDistance  0.007271   0.013507   0.538    0.593    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.1881886)
## 
##     Null deviance: 10.217  on 55  degrees of freedom
## Residual deviance: 10.162  on 54  degrees of freedom
## AIC: 69.347
## 
## Number of Fisher Scoring iterations: 2

## 
## 2017_S1

## 
## Call:
## glm(formula = temp_mean ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -0.98229  -0.25324  -0.06302   0.17760   1.57132  
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        15.76935    0.12832 122.895  < 2e-16 ***
## crossShoreDistance  0.06895    0.01791   3.851 0.000406 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.2917552)
## 
##     Null deviance: 16.288  on 42  degrees of freedom
## Residual deviance: 11.962  on 41  degrees of freedom
## AIC: 73.012
## 
## Number of Fisher Scoring iterations: 2

## 
## 2017_S2

## 
## Call:
## glm(formula = temp_mean ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -0.9385  -0.2437   0.0125   0.3828   0.6618  
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        15.96703    0.09096 175.532   <2e-16 ***
## crossShoreDistance  0.02852    0.01290   2.211   0.0314 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.1764484)
## 
##     Null deviance: 10.0383  on 53  degrees of freedom
## Residual deviance:  9.1753  on 52  degrees of freedom
## AIC: 63.532
## 
## Number of Fisher Scoring iterations: 2

## 
## 2018_S1

## 
## Call:
## glm(formula = temp_mean ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -1.19658  -0.18239   0.03863   0.27564   0.93848  
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        15.273260   0.110813 137.829   <2e-16 ***
## crossShoreDistance  0.008503   0.015627   0.544     0.59    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.1654624)
## 
##     Null deviance: 5.1783  on 32  degrees of freedom
## Residual deviance: 5.1293  on 31  degrees of freedom
## AIC: 38.219
## 
## Number of Fisher Scoring iterations: 2

## 
## 2018_S2

## 
## Call:
## glm(formula = temp_mean ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -1.08258  -0.71334   0.00757   0.70321   1.12396  
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        15.18119    0.21084   72.00   <2e-16 ***
## crossShoreDistance  0.02019    0.02969    0.68    0.502    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.5351738)
## 
##     Null deviance: 14.697  on 28  degrees of freedom
## Residual deviance: 14.450  on 27  degrees of freedom
## AIC: 68.096
## 
## Number of Fisher Scoring iterations: 2

## 
## 2019_S1

## 
## Call:
## glm(formula = temp_mean ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -1.54589  -0.87740   0.06019   0.80505   1.85252  
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        14.95819    0.27068   55.26  < 2e-16 ***
## crossShoreDistance  0.14965    0.03817    3.92 0.000455 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.9872792)
## 
##     Null deviance: 45.780  on 32  degrees of freedom
## Residual deviance: 30.606  on 31  degrees of freedom
## AIC: 97.164
## 
## Number of Fisher Scoring iterations: 2

## 
## 2019_S2

## 
## Call:
## glm(formula = temp_mean ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -1.31068  -0.36402   0.00289   0.27367   1.33878  
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        16.09525    0.15775 102.032   <2e-16 ***
## crossShoreDistance  0.05359    0.02225   2.409   0.0221 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.3353034)
## 
##     Null deviance: 12.340  on 32  degrees of freedom
## Residual deviance: 10.394  on 31  degrees of freedom
## AIC: 61.527
## 
## Number of Fisher Scoring iterations: 2

ggplot(cast.df, aes(crossShoreDistance, temp_mean)) + 
  geom_point() + 
  geom_smooth(method = "glm") +
  facet_wrap('survey')

## `geom_smooth()` using formula 'y ~ x'

<br?

coverage ~ crossShoreDistance

df.split <- split(cast.df, cast.df$survey)
for (i in 1:length(df.split)){
  message('\n', names(df.split)[i])
  fit <- glm(transf_area_sum ~ crossShoreDistance, data=df.split[[i]])
  #print(tab_model.wrap(fit))
  print(summary(fit))
}

## 
## 2015_S1

## 
## Call:
## glm(formula = transf_area_sum ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -0.92939  -0.34872  -0.01466   0.34912   1.48797  
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        -3.449090   0.148364 -23.247   <2e-16 ***
## crossShoreDistance  0.005109   0.018557   0.275    0.785    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.2661982)
## 
##     Null deviance: 8.2723  on 32  degrees of freedom
## Residual deviance: 8.2521  on 31  degrees of freedom
## AIC: 53.911
## 
## Number of Fisher Scoring iterations: 2

## 
## 2016_S1

## 
## Call:
## glm(formula = transf_area_sum ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -1.23263  -0.32272   0.00105   0.32841   1.18628  
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        -3.017598   0.092843 -32.502   <2e-16 ***
## crossShoreDistance -0.002733   0.012668  -0.216     0.83    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.2482193)
## 
##     Null deviance: 17.387  on 71  degrees of freedom
## Residual deviance: 17.375  on 70  degrees of freedom
## AIC: 107.97
## 
## Number of Fisher Scoring iterations: 2

## 
## 2016_S2

## 
## Call:
## glm(formula = transf_area_sum ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -1.05562  -0.32471  -0.01494   0.25733   1.52330  
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        -3.05230    0.10995 -27.761   <2e-16 ***
## crossShoreDistance  0.03217    0.01626   1.978   0.0531 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.2728636)
## 
##     Null deviance: 15.802  on 55  degrees of freedom
## Residual deviance: 14.735  on 54  degrees of freedom
## AIC: 90.153
## 
## Number of Fisher Scoring iterations: 2

## 
## 2017_S1

## 
## Call:
## glm(formula = transf_area_sum ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -0.97148  -0.38227  -0.04024   0.38510   1.13347  
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        -3.06508    0.13610 -22.520   <2e-16 ***
## crossShoreDistance -0.02192    0.01899  -1.154    0.255    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.3282464)
## 
##     Null deviance: 13.895  on 42  degrees of freedom
## Residual deviance: 13.458  on 41  degrees of freedom
## AIC: 78.079
## 
## Number of Fisher Scoring iterations: 2

## 
## 2017_S2

## 
## Call:
## glm(formula = transf_area_sum ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -0.81078  -0.38804   0.06741   0.26360   0.98903  
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        -3.027150   0.098305 -30.794   <2e-16 ***
## crossShoreDistance -0.009255   0.013939  -0.664     0.51    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.2060766)
## 
##     Null deviance: 10.807  on 53  degrees of freedom
## Residual deviance: 10.716  on 52  degrees of freedom
## AIC: 71.914
## 
## Number of Fisher Scoring iterations: 2

## 
## 2018_S1

## 
## Call:
## glm(formula = transf_area_sum ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##     Min       1Q   Median       3Q      Max  
## -1.3107  -0.2820   0.1477   0.3245   0.7379  
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        -2.692090   0.138309 -19.464   <2e-16 ***
## crossShoreDistance -0.006397   0.019505  -0.328    0.745    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.2577605)
## 
##     Null deviance: 8.0183  on 32  degrees of freedom
## Residual deviance: 7.9906  on 31  degrees of freedom
## AIC: 52.848
## 
## Number of Fisher Scoring iterations: 2

## 
## 2018_S2

## 
## Call:
## glm(formula = transf_area_sum ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -0.85044  -0.18479   0.07493   0.25750   0.52566  
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        -2.416415   0.110470 -21.874   <2e-16 ***
## crossShoreDistance -0.002305   0.015556  -0.148    0.883    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.1469135)
## 
##     Null deviance: 3.9699  on 28  degrees of freedom
## Residual deviance: 3.9667  on 27  degrees of freedom
## AIC: 30.607
## 
## Number of Fisher Scoring iterations: 2

## 
## 2019_S1

## 
## Call:
## glm(formula = transf_area_sum ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -0.94351  -0.29033   0.01446   0.28482   0.86312  
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        -3.36498    0.13089 -25.708  < 2e-16 ***
## crossShoreDistance  0.08891    0.01846   4.817 3.63e-05 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.2308488)
## 
##     Null deviance: 12.5128  on 32  degrees of freedom
## Residual deviance:  7.1563  on 31  degrees of freedom
## AIC: 49.209
## 
## Number of Fisher Scoring iterations: 2

## 
## 2019_S2

## 
## Call:
## glm(formula = transf_area_sum ~ crossShoreDistance, data = df.split[[i]])
## 
## Deviance Residuals: 
##      Min        1Q    Median        3Q       Max  
## -0.75516  -0.25987   0.03786   0.32738   0.83785  
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        -3.14945    0.10554 -29.840  < 2e-16 ***
## crossShoreDistance  0.04845    0.01488   3.255  0.00274 ** 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for gaussian family taken to be 0.1501016)
## 
##     Null deviance: 6.2434  on 32  degrees of freedom
## Residual deviance: 4.6531  on 31  degrees of freedom
## AIC: 35.004
## 
## Number of Fisher Scoring iterations: 2

ggplot(cast.df, aes(crossShoreDistance, transf_area_sum)) + 
  geom_point() + 
  geom_smooth(method = "glm") +
  facet_wrap('survey')

## `geom_smooth()` using formula 'y ~ x'