+
+
+
+
+
+
+
+
+
1. Introduction
+
Recall our site suitability exercise in Week 1 - we used our own
+spatial decision factors, or “weights” to impose a rule-based system for
+identifying land areas. We are going to learn an alternative approach,
+where we use a statistical model to algorithmically determine the
+weights and subsequent siting.
+
The use case is as follows - assume you are a program manager at a
+land conservation non-profit, and you are seeking to use your resources
+to acquire land to conserve. Given all the land in Pennsylvania, it’s
+hard know where to start, so you need to find a way to identify suitable
+areas to investigate for acquisition. By using a predictive model, you
+can identify likely lands and begin to research their ownership, and
+determine whether they are available.
+
This is a relatively simplified exercise - there are some detailed
+ways to “tune” models to make them more powerful and flexible - to
+introduce you to logistic regression as a tool for classification. You
+will be able to use this general workflow in your midterm project, when
+you will create your own features for a similar model.
+
+
1.1. Learning objectives
+
+Use binomial logistic regression to create a classification
+prediction.
Use training and testing data to tune a model.
Learn about goodness of fit and error metrics for classification
+models.
Learn about the properties and uses of long data.
Understand the methods you will be using on Assignment 3 and
+understand the predictive modeling workflow.
+
+
+
+
2. Setup
+
To get started, let’s install libraries and a set of ggplot styles we
+call mapTheme and plotTheme.
+
If you don’t have particular libraries installed, run the command
+install.packages('name_of_package_in_quotes') with the
+relevant package, and then call them into your environment with a
+library command.
+
library(caret)
+library(pscl)
+library(plotROC)
+library(pROC)
+library(sf)
+library(tidyverse)
+library(knitr)
+library(kableExtra)
+library(tigris)
+library(viridis)
+
mapTheme <- theme(plot.title =element_text(size=12),
+ plot.subtitle = element_text(size=8),
+ plot.caption = element_text(size = 6),
+ axis.line=element_blank(),
+ axis.text.x=element_blank(),
+ axis.text.y=element_blank(),
+ axis.ticks=element_blank(),
+ axis.title.x=element_blank(),
+ axis.title.y=element_blank(),
+ panel.background=element_blank(),
+ panel.border=element_blank(),
+ panel.grid.major=element_line(colour = 'transparent'),
+ panel.grid.minor=element_blank(),
+ legend.direction = "vertical",
+ legend.position = "right",
+ plot.margin = margin(1, 1, 1, 1, 'cm'),
+ legend.key.height = unit(1, "cm"), legend.key.width = unit(0.2, "cm"))
+
+plotTheme <- theme(
+ plot.title =element_text(size=12),
+ plot.subtitle = element_text(size=8),
+ plot.caption = element_text(size = 6),
+ axis.text.x = element_text(size = 10, angle = 45, hjust = 1),
+ axis.text.y = element_text(size = 10),
+ axis.title.y = element_text(size = 10),
+ # Set the entire chart region to blank
+ panel.background=element_blank(),
+ plot.background=element_blank(),
+ #panel.border=element_rect(colour="#F0F0F0"),
+ # Format the grid
+ panel.grid.major=element_line(colour="#D0D0D0",size=.75),
+ axis.ticks=element_blank())
+
## Warning: The `size` argument of `element_line()` is deprecated as of ggplot2 3.4.0.
+## ℹ Please use the `linewidth` argument instead.
+## This warning is displayed once every 8 hours.
+## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
+## generated.
+
Let’s load our data. We are going to include two datasets:
+preserve, which is a fishnet dataset representing preserved
+land, and protected, which is the original protected lands
+shapefile.
+
How do you make a fishnet on your own? You can learn about that in
+the Intro
+to R course text.
+
preserve <- st_read("https://raw.githubusercontent.com/mafichman/CPLN_675/main/Week_7_10/data/pa_conservation/fishnet3k_pa_JoinDEM_Slope_distSlope_distUrban_landCover_distRivers.geojson")
+
+protected <- st_read("https://raw.githubusercontent.com/mafichman/CPLN_675/main/Week_7_10/data/pa_conservation/pa_protected_lands2.geojson")
+
We are going to load data in that are projected in WGS 84, which is
+the “web mercator” coordinate system, which is in units of decimal
+degrees (e.g. lat/lon). We want to re-project it to something with a
+linear unit in feet or meters so that we can do various distance
+measurements later on.
+
Let’s project it to PA State Plane, which has a CRS (coordinate
+reference system) of 2272.
+
Check out spatialreference.org for
+the crs numbers of all the coordinate systems out there that you might
+want to use.
+
preserve <- preserve %>%
+ st_transform(crs = 2272)
+
+protected <- protected %>%
+ st_transform(crs = 2272)
+
Let’s also load a shape of PA counties from the US Census via the
+tigris package, make sure it’s an sf object
+(st_as_sf) and reproject it to crs = 2272 like our other
+data (st_transform).
+
counties <- counties('PA') %>%
+ st_as_sf() %>%
+ st_transform(crs = 2272)
+
+
+
3. Exploratory analysis
+
Let’s start by checking out the variables we have in our datasets
+using the glimpse command.
+
glimpse(preserve)
+
+glimpse(protected)
+
+
3.1. Maps
+
Let’s plot the sf object of our protected lands on top of our
+counties. We give the shape some color and transparency (alpha).
+
ggplot() +
+ geom_sf(data = counties)+
+ geom_sf(data=protected,
+ fill = "dark green",
+ color = "dark green",
+ alpha = 0.6) +
+ labs(title="Protected lands in Pennsylvania") +
+ mapTheme
+
+
Now let’s plot the fishnet version.
+
Notice we set the fill of our geom_sf to
+as.factor(preserve) and set the color to “transparent”
+outside the aesthetics. What does this do?
+
Can you play with this ggplot and switch the colors and styling?
+
ggplot() +
+ geom_sf(data=preserve, aes(fill=as.factor(preserve)), color = "transparent") +
+ geom_sf(data = counties, fill = "transparent", color = "white")+
+ scale_fill_manual(values = c("dark blue", "dark green"),
+ labels = c("Not Preserved","Preserved"),
+ name = "") +
+ labs(title="Protected lands in Pennsylvania (Fishnet)") +
+ mapTheme
+
+
+
+
3.2. Plots
+
Let’s build some bar plots that show differences in our independent
+variables across land that has and has not been preserved.
+
Notice the use of the gather function. What is it doing?
+Use glimpse to examine the data.
+
preservePlotVariables <-
+ preserve %>%
+ as.data.frame() %>%
+ select(preserve,elevation,slope,dSteepSlop,dUrban,distRivers) %>%
+ gather(variable, value, -preserve)
+
Note there is a more modern version of gather
+known as pivot_longer, where the syntax would look like
+this:
+
preserve %>% as.data.frame() %>%
+ select(preserve,elevation,slope,dSteepSlop,dUrban,distRivers) %>%
+ pivot_longer(cols = -preserve)
+
We are seeing this weird concept of wide and long data again - this
+graphic is pretty helpful understanding how data can transform from one
+form to another.
+
Can you think of a few situations in which one form or another is
+appropriate?
+
Image: https://www.garrickadenbuie.com/
+
Let’s examine some of the variables and how they vary across our
+preserved/not preserved variable. Our preservePlotVariables
+data frame has rows for each cell-variable pair, we
+group_by variable and preservation status and take the mean
+value for each variable by status.
+
ggplot(preservePlotVariables %>%
+ group_by(preserve, variable) %>%
+ summarize(mean = mean(value))) +
+ geom_bar(aes(as.factor(preserve),
+ mean,
+ fill=as.factor(preserve)),
+ stat="identity") +
+ facet_wrap(~variable) +
+ scale_fill_manual(values = c("dark blue", "dark green"),
+ labels = c("Not Preserved","Preserved"),
+ name = "") +
+ labs(x="Preserved", y="Value")
+
## `summarise()` has grouped output by 'preserve'. You can override using the
+## `.groups` argument.
+
+
Question:
+
Can you calculate how many fishnet grid cells are preserved?
+What is it on a percentage basis? Can you re-scale slope to look at it
+how it varies from preserved/not-preserved using the ‘scales’ argument
+in the facet argument?
+
Bonus - check out this code - another way to examine these data,
+what do you see here that might inform a model?
+
ggplot(preservePlotVariables) +
+ geom_violin(aes(x = as.factor(preserve),
+ y = value, fill = as.factor(preserve))) +
+ facet_wrap(~variable, scales = "free") +
+ labs(x="Preserved", y="Value") +
+ scale_fill_manual(values = c("dark blue", "dark green"),
+ labels = c("Not Preserved","Preserved"), name = "") +
+ labs(x="Preserved", y="Value") +
+ plotTheme
+
+
+
+
4. Data wrangling
+
Lets build another data set of just the variables we want to analyze.
+Notice that land cover is an integer but we want it to be a factor, so
+it gets re-coded.
+
preserve <-
+ preserve %>%
+ select(preserve,elevation,slope,dSteepSlop,landCover,dUrban,distRivers, Id) %>%
+ mutate(landCover = as.factor(landCover))
+
+
+
5. Model building
+
In order to assess our model’s accuracy, we are going to split our
+data into training and test sets. The test set will be put aside as we
+train our model on the training set. We then predict for
+the test set values using our model and compare the accuracy of our
+predictions to the known values in the test set to see how well our
+model performs.
+
+
5.1. Partition training and test sets
+
Now we create training and test sets.
+
Let’s look over this operation:
+
+
Why do you think preserve$landCover is specified
+below?
+
Why is it important to have separate training and test
+sets?
+
set.seed(3456)
+trainIndex <- createDataPartition(preserve$landCover, p = .70,
+ list = FALSE,
+ times = 1)
+
## Warning in createDataPartition(preserve$landCover, p = 0.7, list = FALSE, :
+## Some classes have a single record ( 3 ) and these will be selected for the
+## sample
+
preserveTrain <- preserve[ trainIndex,]
+preserveTest <- preserve[-trainIndex,]
+
+
+
5.2. Make a binomial model
+
Now let’s estimate a logistic regression model. The binomial logit
+model runs in the glm function (generalized linear models).
+We specify the dependent variable as preserve and run the
+model on our training set preserveTrain.
+
Note how we can use the dplyr pipes right in the data parameter. We
+have to convert to a data frame because R won’t know how to run a
+regression on an sf.
+
Let’s look at the model output, we see that we have coefficients, and
+p-values, but no R-squared. There are other goodness of fit metrics we
+will look at. The AIC, though not on a 0-1 scale like R-squared, has a
+similar function in that it tells you about overall model fit, but not
+about error and accuracy.
+
We are not really interested in our coefficients other than their
+magnitude, directionality and p-value (generall). But for the record,
+the way the coefficients in a logistic regression are interpreted is
+different than in OLS - we are talking in terms of “odds” of an outcome
+occurring (in our case odds of land being preserved.). If we
+exponentiate the coefficient (exp()) we can interpret it as
+all else equal the exponentiated value being the increase or
+decrease in the odds of the outcome.
+
preserveModel <- glm(preserve ~ .,
+ family="binomial"(link="logit"), data = preserveTrain %>%
+ as.data.frame() %>%
+ select(-geometry, -Id))
+summary(preserveModel)
+
##
+## Call:
+## glm(formula = preserve ~ ., family = binomial(link = "logit"),
+## data = preserveTrain %>% as.data.frame() %>% select(-geometry,
+## -Id))
+##
+## Coefficients:
+## Estimate Std. Error z value Pr(>|z|)
+## (Intercept) -18.606424174 168.144517528 -0.111 0.912
+## elevation 0.002907342 0.000258183 11.261 <0.0000000000000002 ***
+## slope 0.146417480 0.011893467 12.311 <0.0000000000000002 ***
+## dSteepSlop 0.000013548 0.000010373 1.306 0.192
+## landCover1 14.797548603 168.144990634 0.088 0.930
+## landCover2 13.569395402 168.144676533 0.081 0.936
+## landCover3 29.470186539 898.614723616 0.033 0.974
+## landCover4 13.885565593 168.144433869 0.083 0.934
+## landCover5 12.839862237 168.144494025 0.076 0.939
+## landCover6 17.089528172 168.146647130 0.102 0.919
+## dUrban 0.000660267 0.000033638 19.629 <0.0000000000000002 ***
+## distRivers 0.000001583 0.000011459 0.138 0.890
+## ---
+## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
+##
+## (Dispersion parameter for binomial family taken to be 1)
+##
+## Null deviance: 8161.8 on 9339 degrees of freedom
+## Residual deviance: 6416.2 on 9328 degrees of freedom
+## AIC: 6440.2
+##
+## Number of Fisher Scoring iterations: 13
+
+
+
5.3. Model validation
+
Using the predict function, we create a vector of
+classification probabilities we call classProbs. These are
+the predicted probability of a test set (preserveTest)
+fishnet cell being conserved conditional on our model. Setting the
+parameter type="reponse" returns probabilities that range
+from 0 to 1.
+
What do you make of the distribution of classProbs?
+
classProbs <- predict(preserveModel, preserveTest, type="response")
+
+hist(classProbs)
+
+
Let’s put classProbs into a data frame along with the
+observed preserve outome, which is either 1
+for preserved land or 0 for unpreserved.
+
Then we build this funky plot, testProbsPlot. The
+vertical line represents a 0.5 probability of preservation.
+
Can you interperet this plot? Can you come up with a title for
+this plot?
+
testProbs <- data.frame(obs = as.numeric(preserveTest$preserve),
+ pred = classProbs)
+
+ggplot(testProbs, aes(x = pred, fill=as.factor(obs))) +
+ geom_density() +
+ facet_grid(obs ~ .) +
+ xlab("Probability") +
+ ylab("Frequency")+
+ geom_vline(xintercept = .5) +
+ scale_fill_manual(values = c("dark blue", "dark green"),
+ labels = c("Not Preserved","Preserved"),
+ name = "")+
+ plotTheme
+
+
+
5.3.1 Confusion metrics
+
Now we have to figure out at which probability level do we wish
+to classify land as being preserved. How do we make this
+decision?
+
Let’s (arbitrarily for now) choose 50% and then create a table of our
+correct and incorrect predictions, called a “confusion matrix”. Below we
+set the reference to the observed preserved status, data to the
+predicted outcome, and make sure to state which factor level is the
+positive (ie. preserved) level. Note that confusionMatrix
+does not take numeric inputs, only factors.
+
testProbs$predClass = ifelse(testProbs$pred > .5 ,1,0)
+
+caret::confusionMatrix(reference = as.factor(testProbs$obs),
+ data = as.factor(testProbs$predClass),
+ positive = "1")
+
## Confusion Matrix and Statistics
+##
+## Reference
+## Prediction 0 1
+## 0 3266 513
+## 1 75 145
+##
+## Accuracy : 0.853
+## 95% CI : (0.8416, 0.8638)
+## No Information Rate : 0.8355
+## P-Value [Acc > NIR] : 0.001335
+##
+## Kappa : 0.2701
+##
+## Mcnemar's Test P-Value : < 0.00000000000000022
+##
+## Sensitivity : 0.22036
+## Specificity : 0.97755
+## Pos Pred Value : 0.65909
+## Neg Pred Value : 0.86425
+## Prevalence : 0.16454
+## Detection Rate : 0.03626
+## Detection Prevalence : 0.05501
+## Balanced Accuracy : 0.59896
+##
+## 'Positive' Class : 1
+##
+
What is the sensitivy and specificity suggest about our model? What
+is accuracy suggest? Why would we choose a higher threshold cutoff?
+
Predicted = 0, Observed = 0 —> True Negative
+
Predicted = 1, Observed = 1 —> True Positive
+
Predicted = 1, Observed = 0 —> False Positive
+
Predicted = 0, Observed = 1 —> False Negative
+
1. Sensitivity - the proportion of actual positives (1’s)
+that were predicted to be positive. Also known as “true positive
+rate”.
+
2. Specificity - The proportion of actual negatives (0’s)
+that were predicted to be negatives. Also known as “true negative
+rate”.
+
+
+
5.3.2. ROC Curve
+
Let’s create an ROC (receiver operating characteristic) curve. What
+does this tell us?
+
See Appendix 1 for more on ROC curves.
+
ggplot(testProbs, aes(d = obs, m = pred)) +
+ geom_roc(n.cuts = 50, labels = FALSE) +
+ style_roc(theme = theme_grey) +
+ geom_abline(slope = 1, intercept = 0, size = 1.5, color = 'grey')
+
+
How about the area under the curve?
+
auc(testProbs$obs, testProbs$pred)
+
## Setting levels: control = 0, case = 1
+
## Setting direction: controls < cases
+
## Area under the curve: 0.8067
+
+
+
5.3.3. Cross validation
+
Testing the power of your model on out of sample data is critical to
+the machine learning process. Cross-validation iteratively creates many
+randomly generated test sets or ‘folds’, testing the power of your model
+on each.
+
First we set the ctrl parameter which specifies the flavor of cross
+validation we wish to use. You can see all the different cross
+validation options here. In this instance, number = 100 tell us that we
+are going to iteratively test our models on 100 hold out test sets.
+
Then we estimate our model using the train function from the caret
+package. Note that we use the entire preserve data set as the process of
+cross-validation will create test sets for us.
+
Make sure that you update the regression to the model you specified
+above.
+
ctrl <- trainControl(method = "cv",
+ number = 100,
+ p = 0.7,
+ savePredictions = TRUE)
+
+cvFit <- train(as.factor(preserve) ~ ., data = preserve %>%
+ as.data.frame() %>%
+ select(-geometry, -Id),
+ method="glm", family="binomial",
+ trControl = ctrl)
+
+cvFit
+
## Generalized Linear Model
+##
+## 13339 samples
+## 6 predictor
+## 2 classes: '0', '1'
+##
+## No pre-processing
+## Resampling: Cross-Validated (100 fold)
+## Summary of sample sizes: 13205, 13206, 13206, 13205, 13206, 13206, ...
+## Resampling results:
+##
+## Accuracy Kappa
+## 0.8562114 0.276218
+
Notice that the accuracy metric is actually the average accuracy
+across all 100 folds. While that is useful, what we are really
+interested in is the variability of accuracy across all 100 folds.
+Before going any further into that, let’s plot a histogram of accuracy
+across all 100 folds.
+
Before doing so, check out cvFit$resample. What
+information is stored here?
+
ggplot(as.data.frame(cvFit$resample), aes(Accuracy)) +
+ geom_histogram() +
+ scale_x_continuous(limits = c(0, 1)) +
+ labs(x="Accuracy",
+ y="Count")+
+ plotTheme
+
+
The key to this plot is as follows. We want a model that is
+generalizable, particularly to data it hasn’t seen already.
+Cross-validation helps us understand how the model works in this
+context. If our model was not generalizable to ‘out-of-sample’ data,
+then we should expect wildly different accuracies across each of the 100
+folds.
+
Is that what we see here? Do you think we have a generalizable
+model?
+
What if we wanted to know whether our model generalized well across
+different counties (or kinds of counties)?
+
+
+
5.3.2. Map predictions
+
Now that we have tuned our model, let’s predict for the entire
+dataset and assess our predictions.
+
allPredictions <-
+ predict(cvFit, preserve, type="prob")[,2]
+
+preserve <-
+ cbind(preserve,allPredictions) %>%
+ mutate(allPredictions = round(allPredictions * 100))
+
Now we map the predictions.
+
What would you title this map?
+
Note how we use ntile in the aes parameter to create quintiles. (The
+quintile labels are created in the scale_fill_manual function.)
+
ggplot() +
+ geom_sf(data=preserve, aes(fill=factor(ntile(allPredictions,5))),
+ colour=NA) +
+ scale_fill_manual(values = c("#edf8fb","#b3cde3","#8c96c6","#8856a7","#810f7c"),
+ labels=as.character(quantile(preserve$allPredictions,
+ c(0.1,.2,.4,.6,.8),
+ na.rm=T)),
+ name="Predicted\nProbabilities(%)\n(Quintile\nBreaks)") +
+ mapTheme +
+ labs(title="")
+
+
Let’s map it again with the already other land cover types
+overlaid.
+
ggplot() +
+ geom_sf(data=preserve, aes(fill=factor(ntile(allPredictions,5))), colour=NA) +
+ scale_fill_manual(values = c("#edf8fb","#b3cde3","#8c96c6","#8856a7","#810f7c"),
+ labels=as.character(quantile(preserve$allPredictions,
+ c(0.1,.2,.4,.6,.8),
+ na.rm=T)),
+ name="Predicted\nProbabilities(%)\n(Quintile\nBreaks)") +
+ geom_sf(data=preserve %>%
+ filter(preserve == 1),
+ fill="dark green",colour=NA) +
+ geom_sf(data=preserve %>%
+ filter(landCover == 2),
+ fill="red",colour=NA) +
+ mapTheme +
+ labs(title="Observed and Predicted Conservation Areas",
+ subtitle="Pennsylvania; Existing conserved land in green; Existing development in red ")
+
+
We could assess many things about our model by exploring our
+errors.
+
Let’s map our confusion metrics across our entire data set for a 50%
+threshold. Notice I’m doing some dplyr commands to
+preserve and piping it into my ggplot call to keep from
+making a new data object. Nice shortcut.
+
Do we think there’s something systematic to these errors? How
+would we find that out?
+
preserve %>%
+ mutate(confResult=case_when(allPredictions < 50 & preserve==0 ~ "True_Negative",
+ allPredictions >= 50 & preserve==1 ~ "True_Positive",
+ allPredictions < 50 & preserve==1 ~ "False_Negative",
+ allPredictions >= 50 & preserve==0 ~ "False_Positive")) %>%
+ ggplot()+
+ geom_sf(aes(fill = confResult), color = "transparent")+
+ scale_fill_manual(values = c("Red","Orange","Light Blue","Light Green"),
+ name="Outcomes")+
+ labs(title="Confusion Metrics") +
+ mapTheme
+
+
Does this type of model have more, less or the same level of
+usefulness as a raster-based site suitability? Why?
+
+
+
+
+
Appendix 1: ROC Curves - Is it a bird or a German Bomber?
+
In the early days of radar, it was difficult for radar technology to
+tell the difference between incoming bomber aircraft and a flock of
+geese. This was a particular problem in World War 2 as the German
+bombing of London was a daily threat.
+
Imagine we conducted some experiments with a radar, where for 6
+different radar configurations, we counted the rate of true positives
+(ie. classified correctly) for both geese and for german bombers.
+
We also collect data on false positives - the rate of incorrectly
+identified flocks of geese.
+
Below we can see the results for our 6 radar configurations. We want
+Sensitivity to refer to the rate of bombers correctly classified (“True
+Positives”) and Specificity to refer to the number of geese correctly
+classified (“True Negatives”).
+
a <- c(1,0,100,0)
+b <- c(2,35,95,5)
+c <- c(3,60,85,15)
+d <- c(4,85,70,30)
+e <- c(5,92,30,70)
+f <- c(6,100,0,100)
+
+radarObservations <- data.frame(rbind(a,b,c,d,e,f))
+colnames(radarObservations) <- c("Radar_Setting",
+ "Sensitivity_Pct_Planes_Detected",
+ "Specificity_Pct_Geese_Detected",
+ "Pct_of_Geese_Incorrectly_Identified")
+
+kable(radarObservations)
+
+
+
+|
+ |
+
+Radar_Setting
+ |
+
+Sensitivity_Pct_Planes_Detected
+ |
+
+Specificity_Pct_Geese_Detected
+ |
+
+Pct_of_Geese_Incorrectly_Identified
+ |
+
+
+
+
+|
+a
+ |
+
+1
+ |
+
+0
+ |
+
+100
+ |
+
+0
+ |
+
+
+|
+b
+ |
+
+2
+ |
+
+35
+ |
+
+95
+ |
+
+5
+ |
+
+
+|
+c
+ |
+
+3
+ |
+
+60
+ |
+
+85
+ |
+
+15
+ |
+
+
+|
+d
+ |
+
+4
+ |
+
+85
+ |
+
+70
+ |
+
+30
+ |
+
+
+|
+e
+ |
+
+5
+ |
+
+92
+ |
+
+30
+ |
+
+70
+ |
+
+
+|
+f
+ |
+
+6
+ |
+
+100
+ |
+
+0
+ |
+
+100
+ |
+
+
+
+
What can you say about the tradeoff between detecting both bombers
+and geese as we crank up the radar setting?
+
Which radar setting would you choose?
+
Let’s create a plot of True Postive Rate and False Positive Rate. It
+may be easier to consider this relationship as ‘hit rate’ and ‘false
+alarm rate’ respectively.
+
ggplot(radarObservations,
+ aes(100 - Specificity_Pct_Geese_Detected,
+ Sensitivity_Pct_Planes_Detected)) +
+ geom_abline(slope = 1, intercept = 0, size = 1.5, color = 'grey') +
+ geom_point(size=2) +
+ geom_line() +
+ xlab("False positive rate") + ylab("True positive rate")
+
+
Does this plot help you to better understand which radar
+configuration is best? It’s all about tradeoffs.
+
What do you think the grey line represents?
+
We call this kind of plot an “ROC Curve” which stands for Receiver
+Operating Characteristic.
+
What would a reading in the top most left corner represent?
+
+
+
Appendix 2: Measure nearest neighbor distance
+
Imagine that we believe conservation is really a function of distance
+to Unicorn Farms and we wish to engineer a feature that was distance
+from each fishnet grid cell to the nearest unicorn farm. Note that we
+have these instructions in Arc as part of the Week 6 Markdown.
+
First we have to create our unicorn farms. We do this by taking a
+random sample of 50 preserve fishnet grid cells, taking their centroids
+with st_centroid and mapping the mapping them.
+
unicornFarms <-
+ preserve %>%
+ sample_n(50) %>%
+ st_centroid() %>%
+ mutate(legendItem="Unicorn Farms")
+
## Warning: st_centroid assumes attributes are constant over geometries
+
ggplot() +
+ geom_sf(data=st_sf(st_union(preserve))) +
+ geom_point(data=unicornFarms, aes(st_coordinates(unicornFarms)[,1],
+ st_coordinates(unicornFarms)[,2],
+ colour="Unicorn Farms")) +
+ mapTheme
+
+
Next, in order to measure ‘nearest neighbor distance’, we have to
+convert both the unicorm farms and the preserve fishnet to matrices of
+xy centroid coordinates like so:
+
preserveXY <-
+ preserve %>%
+ st_centroid %>%
+ st_coordinates %>%
+ as.matrix()
+
## Warning: st_centroid assumes attributes are constant over geometries
+
unicornXY <-
+ unicornFarms %>%
+ st_coordinates %>%
+ as.matrix()
+
+head(unicornXY)
+
## X Y
+## [1,] 1583707 244268.6
+## [2,] 1442824 348091.2
+## [3,] 2662962 681559.6
+## [4,] 1269093 244436.8
+## [5,] 2537155 923937.1
+## [6,] 1307460 275097.6
+
Next we can use the below ‘k nearest neighbor’ function to measure
+the average nearest neighbor distance from each preserve fishnet
+centroid to its nearest 1 (ie k=1) unicorn farm neighbor.
+
It would be really helpful for you to run through this function line
+by line so that you understand what is going on.
+
library(FNN)
+
+nn_function <- function(measureFrom,measureTo,k) {
+
+nn <-
+ get.knnx(measureTo, measureFrom, k)$nn.dist
+
+output <-
+ as.data.frame(nn) %>%
+ rownames_to_column(var = "thisPoint") %>%
+ gather(points, point_distance, V1:ncol(.)) %>%
+ arrange(as.numeric(thisPoint)) %>%
+ group_by(thisPoint) %>%
+ summarize(pointDistance = mean(point_distance)) %>%
+ arrange(as.numeric(thisPoint)) %>%
+ dplyr::select(-thisPoint)
+
+return(output)
+ }
+
Now let’s run the function, append unicornDistance to our fishnet
+centroids data frame and map. Notice, I bind the result to thepreserve
+data frame with bind_cols.
+
See what’s going on with the nearest neighbor distance?
+
unicornDistance <- nn_function(preserveXY, unicornXY, 1)
+
+output <- bind_cols(preserve, unicornDistance)
+
+ggplot() +
+ geom_sf(data=st_sf(st_union(preserve))) +
+ geom_sf(data=output, aes(fill=pointDistance),colour=NA) +
+ geom_point(data=unicornFarms, aes(st_coordinates(unicornFarms)[,1],
+ st_coordinates(unicornFarms)[,2])) +
+ mapTheme
+
+
One nearest neighbor may be fine, but what if we want to increase the
+scale and measure average distance to three nearest unicorn farms?
+
unicornDistance <- nn_function(preserveXY, unicornXY, 3)
+
+output <- bind_cols(preserve, unicornDistance)
+
+ggplot() +
+ geom_sf(data=st_sf(st_union(preserve))) +
+ geom_sf(data=output, aes(fill=pointDistance),colour=NA) +
+ geom_point(data=unicornFarms, aes(st_coordinates(unicornFarms)[,1],
+ st_coordinates(unicornFarms)[,2])) +
+ mapTheme
+
+
+
+
Appendix 3. Geoprocessing and data-wrangling vignettes
+
In order to make other spatial variables for models you are building
+in R, you might find yourself wanting to summarize vector or raster
+information inside a fishnet.
+
Let’s take our basic fishnet preserve from the earlier
+part of this exercise, and think about how we are going to wrangle some
+basic data into the fishnet and engineer some features.
+
+
Spatial joins and summaries
+
Let’s join and summarize some spatial data to get it into our
+fishnet.
+
First let’s see which state forest district each cell belongs to (these
+data come from PASDA and PA DCNR). We st_read state
+forest districts, reproject them to crs = 2272 so they
+match preserve (not sure what CRS your data are in? query
+them with st_crs)
+
forest_districts <- st_read("https://raw.githubusercontent.com/mafichman/CPLN_675/main/Week_7_10/data/pa_conservation/DCNR_BOF_Bndry_SFM201703.geojson") %>%
+ st_transform(2272)
+
We can quickly check the data and see how they look. They are large
+polygons.
+
ggplot()+
+ geom_sf(data = forest_districts)
+
+
We can join our preserve centroids to these polygons.
+(Why are we joining the centroids to the polygons? Think about the scale
+biases associated with different types of joins.)
+
We will do this in a few steps, but it’s easiest when it’s all
+“piped” together.
+
# Turn preserve into centroid points and join
+
+preserve_and_forests_centroid <-
+ st_join(preserve %>%
+ st_centroid(),
+ forest_districts)
+
+# Turn this data into a data frame, throw away everything but the new data and the unique id (select) of the fishnet cells and then join it back go the original
+
+preserve_and_forests_fishnet <-
+ left_join(preserve,
+ preserve_and_forests_centroid %>%
+ as.data.frame() %>%
+ select(DistrictNa, Id))
+
Now let’s load in some points representing local park access points,
+and summarize how many points we find in each of our fishnet cells.
+These data also
+come from DCNR, and we reproject them to crs = 2272 as
+well.
+
local_parks <- st_read("https://raw.githubusercontent.com/mafichman/CPLN_675/main/Week_7_10/data/pa_conservation/DCNR_LocalParkAccess201511.geojson") %>%
+ st_transform(2272)
+
Inspect the data and you will see that these are thousands of points
+across the commonwealth. Let’s find out how many we have in each of our
+cells, and also calculate how many we have by unit area.
+
Notice we join the parks on the left hand side of this join. What
+happens if you do it the other way?
+
parks_and_fishnet <- st_join(local_parks, preserve_and_forests_fishnet)
+
You can look at these data and see that we have lots of extraneous
+info. What do we really care about here? Number of points per cell. So
+let’s group by our unique identifier, Id and summarize (you
+can use this n() way of doing things, or there are
+alternatives like tally), then join this summary back to
+the fishnet… all at once!
+
Here we do a ’right_join` - we keep all the observations on the right
+hand side (e.g. our fishnet cells), and only the things on the left that
+we need.
+
What is the . thing you ask? Well in a join you need two
+sides to write the function - so the period stands in for the data being
+piped in.
+
preserve_and_forests_and_parks_fishnet <-
+ parks_and_fishnet %>%
+ as.data.frame() %>% # Need to make the sf a dataframe to throw away the geometry
+ group_by(Id) %>%
+ summarize(n_parks = n()) %>%
+ right_join(., preserve_and_forests_fishnet) %>%
+ st_as_sf()
+
Now let’s do an area normalization calculation and create a count of
+points by area. Mind the projection. Remember we are in
+crs = 2272 that’s a projected coordinate system with a
+linear unit in feet. If you are using WGS84 aka
+crs 4326 aka lat /lon aka
+google maps projection this will give you a calculation in
+decimal degrees, which is not good.
+
The key is to use the st_area - and then we get a
+measurement in the linear unit.
+
We create parks_sqmi by dividing the number of parks in
+the cell by the area (transformed into miles from feet). I looked up on
+google that 1 square foot is 3.587e-8 square miles.
+
preserve_and_forests_and_parks_fishnet <-
+ preserve_and_forests_and_parks_fishnet %>%
+ mutate(area = st_area(geometry),
+ parks_sqmi = n_parks/(as.numeric(area)*.00000003587))
+
Now I’m going to make a map just to check it all out:
+
ggplot()+
+ geom_sf(data = preserve_and_forests_and_parks_fishnet,
+ aes(fill = parks_sqmi), color = "transparent")+
+ scale_fill_viridis()+
+ labs(title="Density of park access points",
+ subtitle="Points/ sqmi ")+
+ mapTheme
+
+
+
+
Rescale/recategorize data
+
What if I wasn’t so keen on steep slopes being related in a linear
+way to our outcome? Maybe I can reclassify them into what we call a
+“dummy variable” (aka a “fixed effect”) - what if we call some new
+levels “high slope”, “medium slope” and “low slope.” I can do this with
+a case_when for multiple categories. I could also use
+ifelse to create two categories. Check out these
+examples:
+
preserve_and_forests_and_parks_fishnet %>%
+ mutate(slope_category_2 = ifelse(slope >= 6, "high slope", "low slope"),
+ slope_category_3 = case_when(slope < 3 ~ "low slope",
+ slope >= 3 & slope <6 ~ "medium slope",
+ slope >= 6 ~ "high slope"))
+
## Simple feature collection with 13339 features and 15 fields
+## Geometry type: POLYGON
+## Dimension: XY
+## Bounding box: xmin: 1183997 ymin: 131792 xmax: 2817775 ymax: 1080121
+## Projected CRS: NAD83 / Pennsylvania South (ftUS)
+## # A tibble: 13,339 × 16
+## Id n_parks preserve elevation slope dSteepSlop landCover dUrban distRivers
+## * <int> <int> <int> <dbl> <dbl> <dbl> <fct> <dbl> <dbl>
+## 1 27 1 0 198 1.56 14834 5 1102 2640
+## 2 34 1 0 402 6.91 1694 4 533 5740
+## 3 37 1 0 167 4.22 1808 4 863 721
+## 4 48 2 0 238 4.54 13293 5 2325 4794
+## 5 56 4 0 265 3.14 8798 5 377 2128
+## 6 57 2 1 256 3.26 10796 5 1032 1958
+## 7 63 1 0 197 3.15 6512 5 5093 3575
+## 8 71 1 1 110 3.11 3964 5 1343 707
+## 9 75 1 0 146 2.73 4196 5 802 1779
+## 10 77 1 0 142 2.11 9757 5 2647 1932
+## # ℹ 13,329 more rows
+## # ℹ 7 more variables: allPredictions <dbl>, DistrictNa <chr>,
+## # geometry <POLYGON [US_survey_foot]>, area [US_survey_foot^2],
+## # parks_sqmi <dbl>, slope_category_2 <chr>, slope_category_3 <chr>
+
+
+
+
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
+
+
+