I am using genetic matching in R using GenMatch in order to find comparable treatment and control groups to estimate a treatment effect. The default code for matching looks as follows:
GenMatch(Tr, X, BalanceMatrix=X, estimand="ATT", M=1, weights=NULL,
pop.size = 100, max.generations=100,...)
The description for the pop.size argument in the package is:
Population Size. This is the number of individuals genoud uses to
solve the optimization problem. The theorems proving that genetic
algorithms find good solutions are asymptotic in population size.
Therefore, it is important that this value not be small. See genoud
for more details.
Looking at gnoud the additional description is:
...There are several restrictions on what the value of this number can
be. No matter what population size the user requests, the number is
automatically adjusted to make certain that the relevant restrictions
are satisfied. These restrictions originate in what is required by
several of the operators. In particular, operators 6 (Simple
Crossover) and 8 (Heuristic Crossover) require an even number of
individuals to work on—i.e., they require two parents. Therefore, the
pop.size variable and the operators sets must be such that these three
operators have an even number of individuals to work with. If this
does not occur, the population size is automatically increased until
this constraint is satisfied.
I want to know how gnoud (resp. GenMatch) incorporates the population size argument. Does the algorithm randomly select n individuals from the population for the optimization?
I had a look at the package description and the source code, but did not find a clear answer.
The word "individuals" here does not refer to individuals in the sample (i.e., individual units your dataset), but rather to virtual individuals that the genetic algorithm uses. These individuals are individual draws of a set of the variables to be optimized. They are unrelated to your sample.
The goal of genetic matching is to choose a set of scaling factors (which the Matching documentation calls weights), one for each covariate, that weight the importance of that covariate in a scaled Euclidean distance match. I'm no expert on the genetic algorithm, but my understanding of what it does is that it makes a bunch of guesses at the optimal values of these scaling factors, keeps the ones that "do the best" in the sense of optimizing the criterion (which is determined by fit.func in GenMatch()), and creates new guesses as slight perturbations of the kept guesses. It then repeats this process many times, simulating what natural selection does to optimize traits in living things. Each guess is what the word "individual" refers to in the description for pop.size, which corresponds to the number of guesses at each generation of the algorithm.
GenMatch() always uses your entire sample (unless you have provided a restriction like a caliper, exact matching requirement, or common support rule); it does not sample units from your sample to form each guess (which is what bagging in is other machine learning contexts).
Results will change over many runs because the genetic algorithm itself is a stochastic process. It may converge to a solution asymptotically, but because it is optimizing over a lumpy surface, it will find different solutions each time in finite samples with finite generations and a finite population size (i.e., pop.size).
Related
I could use some advice on methods in R to determine the optimal number of clusters and later on describe the clusters with different statistical criteria. I’m new to R with basic knowledge about the statistical foundations of cluster analysis.
Methods to determine the number of clusters: In the literature one common method to do so is the so called "Elbow-criterion" which compares the Sum of Squared Differences (SSD) for different cluster solutions. Therefore the SSD is plotted against the numbers of Cluster in the analysis and an optimal number of clusters is determined by identifying the “elbow” in the plot (e.g. here: https://en.wikipedia.org/wiki/File:DataClustering_ElbowCriterion.JPG)
This method is a first approach to get a subjective impression. Therefore I’d like to implement it in R. The information on the internet on this is sparse. There is one good example here: http://www.mattpeeples.net/kmeans.html where the author also did an interesting iterative approach to see if the elbow is somehow stable after several repetitions of the clustering process (nevertheless it is for partitioning cluster methods not for hierarchical).
Other methods in Literature comprise the so called “stopping rules”. MILLIGAN & COOPER compared 30 of these stopping rules in their paper “An examination of procedures for determining the number of clusters in a data set” (available here: http://link.springer.com/article/10.1007%2FBF02294245) finding that the Stopping Rule from Calinski and Harabasz provided the best results in a Monte Carlo evaluation. Information on implementing this in R is even sparser.
So if anyone has ever implemented this or another Stopping rule (or other method) some advice would be very helpful.
Statistically describe the clusters:For describing the clusters I thought of using the mean and some sort of Variance Criterion. My data is on agricultural land-use and shows the production numbers of different crops per Municipality. My aim is to find similar patterns of land-use in my dataset.
I produced a script for a subset of objects to do a first test-run. It looks like this (explanations on the steps within the script, sources below).
#Clusteranalysis agriculture
#Load data
agriculture <-read.table ("C:\\Users\\etc...", header=T,sep=";")
attach(agriculture)
#Define Dataframe to work with
df<-data.frame(agriculture)
#Define a Subset of objects to first test the script
a<-df[1,]
b<-df[2,]
c<-df[3,]
d<-df[4,]
e<-df[5,]
f<-df[6,]
g<-df[7,]
h<-df[8,]
i<-df[9,]
j<-df[10,]
k<-df[11,]
#Bind the objects
aTOk<-rbind(a,b,c,d,e,f,g,h,i,j,k)
#Calculate euclidian distances including only the columns 4 to 24
dist.euklid<-dist(aTOk[,4:24],method="euclidean",diag=TRUE,upper=FALSE, p=2)
print(dist.euklid)
#Cluster with Ward
cluster.ward<-hclust(dist.euklid,method="ward")
#Plot the dendogramm. define Labels with labels=df$Geocode didn't work
plot(cluster.ward, hang = -0.01, cex = 0.7)
#here are missing methods to determine the optimal number of clusters
#Calculate different solutions with different number of clusters
n.cluster<-sapply(2:5, function(n.cluster)table(cutree(cluster.ward,n.cluster)))
n.cluster
#Show the objects within clusters for the three cluster solution
three.cluster<-cutree(cluster.ward,3)
sapply(unique(three.cluster), function(g)aTOk$Geocode[three.cluster==g])
#Calculate some statistics to describe the clusters
three.cluster.median<-aggregate(aTOk[,4:24],list(three.cluster),median)
three.cluster.median
three.cluster.min<-aggregate(aTOk[,4:24],list(three.cluster),min)
three.cluster.min
three.cluster.max<-aggregate(aTOk[,4:24],list(three.cluster),max)
three.cluster.max
#Summary statistics for one variable
three.cluster.summary<-aggregate(aTOk[,4],list(three.cluster),summary)
three.cluster.summary
detach(agriculture)
Sources:
http://www.r-tutor.com/gpu-computing/clustering/distance-matrix
How to apply a hierarchical or k-means cluster analysis using R?
http://statistics.berkeley.edu/classes/s133/Cluster2a.html
The elbow criterion as your links indicated is for k-means. Also the cluster mean is obviously related to k-means, and is not appropriate for linkage clustering (in particular not for single-linkage, see single-link-effect).
Your question title however mentions hierarchical clustering, and so does your code?
Note that the elbow criterion does not choose the optimal number of clusters. It chooses the optimal number of k-means clusters. If you use a different clustering method, it may need a different number of clusters.
There is no such thing as the objectively best clustering. Thus, there also is no objectively best number of clusters. There is a rule of thumb for k-means that chooses a (maybe best) tradeoff between number of clusters and minimizing the target function (because increasing the number of clusters always can improve the target function); but that is mostly to counter a deficit of k-means. It is by no means objective.
Cluster analysis in itself is not an objective task. A clustering may be mathematically good, but useless. A clustering may score much worse mathematically, but it may provide you insight to your data that cannot be measured mathematically.
This is a very late answer and probably not useful for the asker anymore - but maybe for others. Check out the package NbClust. It contains 26 indices that give you a recommended number of clusters (and you can also choose your type of clustering). You can run it in such a way that you get the results for all the indices and then you can basically go with the number of clusters recommended by most indices. And yes, I think the basic statistics are the best way to describe clusters.
You can also try the R-NN Curves method.
http://rguha.net/writing/pres/rnn.pdf
K means Clustering is highly sensitive to the scale of data e.g. for a person's age and salary, if not normalized, K means would consider salary more important variable for clustering rather than age, which you do not want. So before applying the Clustering Algorithm, it is always a good practice to normalize the scale of data, bring them to the same level and then apply the CA.
So I learned about the concept of information entropy from Khan Academy where is was phrased in the form of "average amount of yes or no questions needed per symbol". They also gave an alternative form using logarithms.
So let's say we have a symbol generator that produces A,B, and C.
P(A)=1/2, P(B)=1/3, and P(C)=1/6
According to their method, I would gat a chart like this:
First method
Then I would multiply their probability of occurring by the amount of questions needed for each giving
(1/2)*1+(1/3)*2+(1/6)*2 = 1.5bits
but their other method gives
-(1/2)log2(1/2)-(1/3)log2(1/3)-(1/6)log2(1/6)= 1.459... bits
The difference is small, but still significant. I've tried this with different combinations and probabilities and got similar results. Is there something I'm missing? Am I using either method wrong, or is one of them more conditional?
Your second calculation is correct.
The problem with your decision tree approach is that the decision tree is not optimal (and indeed, no binary decision tree could be for those probabilities). Your “is it B” decision node represents less than one bit of information, since once you get there you already know it’s probably B. So your decision tree represents a potential encoding of symbols which is expected to consume 1.5 bits on average, but it represents slightly less than 1.5 bits of information.
In order to have a binary tree which represents an optimal encoding, each node needs to have balanced probabilities. This is not possible if some symbol has a probability whose denominator is not a power of 2.
I want to perform mean-shift clustering in R and found out that there are at least two packages that have this functionality: MeanShift and meanShiftR. As showed here the latter is much faster and as I tried out the first one and it took a long time to perform a clustering, I'm keen on choosing meanShiftR. However meanShiftR::meanShift function has rather uncommon way of bandwidth specification, see part of documentation:
queryData A matrix or vector of points to be classified by the mean
shift algorithm. Values must be finite and non-missing.
bandwidth A vector of length equal to the number of columns in the queryData matrix, or length one when queryData is a vector. This
value will be used in the kernel density estimate for steepest ascent
classification. The default is one for each dimension.
I'm not an expert in mean-shift clustering, but the only banwidth specifications I have found in the literature is that bandwidth is scalar or positive definite, symmetric matrix, not a vector. So is this the technical trick to represent the bandwidth and the value of bandwidth have to be the same for each dimension? Or maybe it can vary?
The other issue is that even setting the same value of bandwidth in meanShiftR package as in MeanShift::msClustering, but just replicated to match the number of columns, I've obtained totally different results, in particular much larger number of cluster. Also, the modes were rather very similar and not representative of the dataset. That made me wonder if this package works correct. Have someone even used meanShiftR? If so, maybe you could present any example as the documentation is not clear enough for me?
This isn't actually different.
One scalar per query point.
I could use some advice on methods in R to determine the optimal number of clusters and later on describe the clusters with different statistical criteria. I’m new to R with basic knowledge about the statistical foundations of cluster analysis.
Methods to determine the number of clusters: In the literature one common method to do so is the so called "Elbow-criterion" which compares the Sum of Squared Differences (SSD) for different cluster solutions. Therefore the SSD is plotted against the numbers of Cluster in the analysis and an optimal number of clusters is determined by identifying the “elbow” in the plot (e.g. here: https://en.wikipedia.org/wiki/File:DataClustering_ElbowCriterion.JPG)
This method is a first approach to get a subjective impression. Therefore I’d like to implement it in R. The information on the internet on this is sparse. There is one good example here: http://www.mattpeeples.net/kmeans.html where the author also did an interesting iterative approach to see if the elbow is somehow stable after several repetitions of the clustering process (nevertheless it is for partitioning cluster methods not for hierarchical).
Other methods in Literature comprise the so called “stopping rules”. MILLIGAN & COOPER compared 30 of these stopping rules in their paper “An examination of procedures for determining the number of clusters in a data set” (available here: http://link.springer.com/article/10.1007%2FBF02294245) finding that the Stopping Rule from Calinski and Harabasz provided the best results in a Monte Carlo evaluation. Information on implementing this in R is even sparser.
So if anyone has ever implemented this or another Stopping rule (or other method) some advice would be very helpful.
Statistically describe the clusters:For describing the clusters I thought of using the mean and some sort of Variance Criterion. My data is on agricultural land-use and shows the production numbers of different crops per Municipality. My aim is to find similar patterns of land-use in my dataset.
I produced a script for a subset of objects to do a first test-run. It looks like this (explanations on the steps within the script, sources below).
#Clusteranalysis agriculture
#Load data
agriculture <-read.table ("C:\\Users\\etc...", header=T,sep=";")
attach(agriculture)
#Define Dataframe to work with
df<-data.frame(agriculture)
#Define a Subset of objects to first test the script
a<-df[1,]
b<-df[2,]
c<-df[3,]
d<-df[4,]
e<-df[5,]
f<-df[6,]
g<-df[7,]
h<-df[8,]
i<-df[9,]
j<-df[10,]
k<-df[11,]
#Bind the objects
aTOk<-rbind(a,b,c,d,e,f,g,h,i,j,k)
#Calculate euclidian distances including only the columns 4 to 24
dist.euklid<-dist(aTOk[,4:24],method="euclidean",diag=TRUE,upper=FALSE, p=2)
print(dist.euklid)
#Cluster with Ward
cluster.ward<-hclust(dist.euklid,method="ward")
#Plot the dendogramm. define Labels with labels=df$Geocode didn't work
plot(cluster.ward, hang = -0.01, cex = 0.7)
#here are missing methods to determine the optimal number of clusters
#Calculate different solutions with different number of clusters
n.cluster<-sapply(2:5, function(n.cluster)table(cutree(cluster.ward,n.cluster)))
n.cluster
#Show the objects within clusters for the three cluster solution
three.cluster<-cutree(cluster.ward,3)
sapply(unique(three.cluster), function(g)aTOk$Geocode[three.cluster==g])
#Calculate some statistics to describe the clusters
three.cluster.median<-aggregate(aTOk[,4:24],list(three.cluster),median)
three.cluster.median
three.cluster.min<-aggregate(aTOk[,4:24],list(three.cluster),min)
three.cluster.min
three.cluster.max<-aggregate(aTOk[,4:24],list(three.cluster),max)
three.cluster.max
#Summary statistics for one variable
three.cluster.summary<-aggregate(aTOk[,4],list(three.cluster),summary)
three.cluster.summary
detach(agriculture)
Sources:
http://www.r-tutor.com/gpu-computing/clustering/distance-matrix
How to apply a hierarchical or k-means cluster analysis using R?
http://statistics.berkeley.edu/classes/s133/Cluster2a.html
The elbow criterion as your links indicated is for k-means. Also the cluster mean is obviously related to k-means, and is not appropriate for linkage clustering (in particular not for single-linkage, see single-link-effect).
Your question title however mentions hierarchical clustering, and so does your code?
Note that the elbow criterion does not choose the optimal number of clusters. It chooses the optimal number of k-means clusters. If you use a different clustering method, it may need a different number of clusters.
There is no such thing as the objectively best clustering. Thus, there also is no objectively best number of clusters. There is a rule of thumb for k-means that chooses a (maybe best) tradeoff between number of clusters and minimizing the target function (because increasing the number of clusters always can improve the target function); but that is mostly to counter a deficit of k-means. It is by no means objective.
Cluster analysis in itself is not an objective task. A clustering may be mathematically good, but useless. A clustering may score much worse mathematically, but it may provide you insight to your data that cannot be measured mathematically.
This is a very late answer and probably not useful for the asker anymore - but maybe for others. Check out the package NbClust. It contains 26 indices that give you a recommended number of clusters (and you can also choose your type of clustering). You can run it in such a way that you get the results for all the indices and then you can basically go with the number of clusters recommended by most indices. And yes, I think the basic statistics are the best way to describe clusters.
You can also try the R-NN Curves method.
http://rguha.net/writing/pres/rnn.pdf
K means Clustering is highly sensitive to the scale of data e.g. for a person's age and salary, if not normalized, K means would consider salary more important variable for clustering rather than age, which you do not want. So before applying the Clustering Algorithm, it is always a good practice to normalize the scale of data, bring them to the same level and then apply the CA.
I have ran clv package which consists of S_Dbw and SD validity indexes for clustering purposes in R commander. (http://cran.r-project.org/web/packages/clv/index.html)
I evaluated my clustering results from DBSCAN, K-Means, Kohonen algorithms with S_Dbw index. but for all these three algorithms S_Dbw is "Inf".
Is it "Infinite" meaning? Why did i confront with "Inf". Is there any problem in my clustering results?
In general, when is S_Dbw index result "Inf"?
Be careful when comparing different algorithms with such an index.
The reason is that the index is pretty much an algorithm in itself. One particular clustering will necessarily be the "best" for each index. The main difference between an index and an actual clustering algorithm is that the index doesn't tell you how to find the "best" solution.
Some examples: k-means minimizes the distances from cluster members to cluster centers. Single-link hierarchical clustering will find the partition with the optimal minimum distance between partitions. Well, DBSCAN will find the partitioning of the dataset, where all density-connected points are in the same partition. As such, DBSCAN is optimal - if you use the appropriate measure.
Seriously. Do not assume that because one algorithm scores higher than another in a particular measure means that the algorithm works better. All that you find out this way is that a particular algorithm is more (cor-)related to a particular measure. Think of it as a kind of correlation between the measure and the algorithm, on a conceptual level.
Using a measure for comparing different results of the same algorithm is different. Then obviously there shouldn't be a benefit from one algorithm over itself. There might still be a similar effect with respect to parameters. For example the in-cluster distances in k-means obviously should go down when you increase k.
In fact, many of the measures are not even well-defined on DBSCAN results. Because DBSCAN has the concept of noise points, which the indexes do not AFAIK.
Do not assume that the measure will either give you an indication of what is "true" or "correct". And even less, what is useful or new. Because you should be using cluster analysis not to find a mathematical optimum of a particular measure, but to learn something new and useful about your data. Which probably is not some measure number.
Back to the indices. They usually are totally designed around k-means. From a short look at S_Dbw I have the impression that the moment one "cluster" consists of a single object (e.g. a noise object in DBSCAN), the value will become infinity - aka: undefined. It seems as if the authors of that index did not consider this corner case, but only used it on toy data sets where such situations did not arise. The R implementation can't fix this, without diverting from the original index and instead turning it into yet another index. Handling noise objects and singletons is far from trivial. I have not yet seen an index that doesn't fail in one way or another - typically, a solution such as "all objects are noise" will either score perfect, or every clustering can trivially be improved by putting each noise object to the nearest non-singleton cluster. If you want your algorithm to be able to say "this object doesn't belong to any cluster" then I do not know any appropriate index.
The IEEE floating point standard defines Inf and -Inf as positive and negative infinity respectively. It means your result was too large to represent in the given number of bits.