I applied the fitdistrplus package in order to fit an empirical distribution.
It turned out the best fit was the negative binomial distribution with parameters:
size=0.6900788
mu=2.6522087
dnbinom(0:10, mu = 2.6522087, size =0.6900788)
[1] 0.33666338 0.18435650 0.12362301 0.08796440 0.06439416 0.04793144 0.03607044 0.02735574 0.02086667 0.01598815 0.01229390
I am now trying to generate the same numbers on EXCEL where the parameters are required in different format:
NEGBINOMDIST(number_f,number_s,probability_s)
How am I meant to do this? Any ideas?
Many thanks..
According to Microsoft's documentation, Excel uses the standard "number of draws before n failures" definition; the parameterization used by fitdistrplus is the alternative referred to in ?dnbinom as:
An alternative parametrization (often used in ecology) is by the
mean ‘mu’, and ‘size’, the dispersion parameter, where ‘prob’
= ‘size/(size+mu)’. The variance is ‘mu + mu^2/size’ in this
parametrization.
So if you want to get back from mu and size to prob and size (Excel's probability_s and number_s respectively) you need
number_s=size
probability_s=size/(size+mu)
muval <- 2.6522087
sizeval <- 0.6900788
(probval <- sizeval/(sizeval+muval))
## [1] 0.206469
all.equal(dnbinom(0:10,mu=muval,size=sizeval),
dnbinom(0:10,prob=probval,size=sizeval))
## TRUE
However, you're not done yet, because (as commented above by #James) Excel only allows positive integers for number_s, and the estimated value above is 0.69. You may need to search/ask on an Excel-related forum about how to overcome this limitation ... at worst, since Excel does have an implementation of the gamma function, you can use the formula given in ?dnbinom
Gamma(x+n)/(Gamma(n) x!) p^n (1-p)^x
to implement your own calculation of the NB (this formulation allows non-integer values of n). It would be best to use the GAMMLN function in Excel to calculate the numerator and denominator of the normalization constant on the log scale ... if you're lucky, someone out there will have saved you some trouble and implemented this already ...
Related
Background
I want to generate multivariate distributed random numbers with a fixed variance matrix. For example, I want to generate a 2 dimensional data with covariance value = 0.5, each dimensional variance = 1. The first maginal of data is a norm distribution with mean = 0, sd = 1, and the next is a exponential distribution with rate = 2.
My attempt
My attempt is that we can generate a correlated multinormal distribution random numbers and then revised them to be any distribution by Inverse transform sampling.
In below, I give an example about transforming 2 dimensional normal distribution random numbers into a norm(0,1)+ exp(2) random number:
# generate a correlated multi-normal distribution, data[,1] and data[,2] are standard norm
data <- mvrnorm(n = 1000,mu = c(0,0), Sigma = matrix(c(1,0.5,0.5,1),2,2))
# calculate the cdf of dimension 2
exp_cdf = ecdf(data[,2])
Fn = exp_cdf(data[,2])
# inverse transform sampling to get Exponetial distribution with rate = 2
x = -log(1-Fn + 10^(-5))/2
mean(x);cor(data[,1],x)
Out:
[1] 0.5035326
[1] 0.436236
From the outputs, the new x is a set of exponential(rate = 2) random numbers. Also, x and data[,1] are correlated with 0.43. The correlated variance is 0.43, not very close to my original setting value 0.5. It maybe a issue. I think covariance of sample generated should stay more closer to initial setting value. In general, I think my method is not quite decent, maybe you guys have some amazing code snippets.
My question
As a statistics graduate, I know there exist 10+ methods to generate multivariate random numbers theoretically. In this post, I want to collect bunch of code snippets to do it automatically using packages or handy . And then, I will compare them from different aspects, like time consuming and quality of data etc. Any ideas is appreciated!
Note
Some users think I am asking for package recommendation. However, I am not looking for any recommendation. I already knew commonly used statistical theroms and R packages. I just wanna know how to generate multivariate distributed random numbers with a fixed variance matrix decently and give a code example about generate norm + exp random numbers. I think there must exist more powerful code snippets to do it in a decent way! So I ask for help right now!
Sources:
generating-correlated-random-variables, math
use copulas to generate multivariate random numbers, stackoverflow
Ross simulation, theoretical book
R CRAN distribution task View
I try to understand the differences between the two methods bayes and mle in the bn.fit function of the package bnlearn.
I know about the debate between the frequentist and the bayesian approach on understanding probabilities. On a theoretical level I suppose the maximum likelihood estimate mle is a simple frequentist approach setting the relative frequencies as the probability. But what calculations are done to get the bayes estimate? I already checked out the bnlearn documenation, the description of the bn.fit function and some application examples, but nowhere there's a real description of what's happening.
I also tried to understand the function in R by first checking out bnlearn::bn.fit, leading to bnlearn:::bn.fit.backend, leading to bnlearn:::smartSapply but then I got stuck.
Some help would be really appreciated as I use the package for academic work and therefore I should be able to explain what happens.
Bayesian parameter estimation in bnlearn::bn.fit applies to discrete variables. The key is the optional iss argument: "the imaginary sample size used by the bayes method to estimate the conditional probability tables (CPTs) associated with discrete nodes".
So, for a binary root node X in some network, the bayes option in bnlearn::bn.fit returns (Nx + iss / cptsize) / (N + iss) as the probability of X = x, where N is your number of samples, Nx the number of samples with X = x, and cptsize the size of the CPT of X; in this case cptsize = 2. The relevant code is in the bnlearn:::bn.fit.backend.discrete function, in particular the line: tab = tab + extra.args$iss/prod(dim(tab))
Thus, iss / cptsize is the number of imaginary observations for each entry in a CPT, as opposed to N, the number of 'real' observations. With iss = 0 you would be getting a maximum likelihood estimate, as you would have no prior imaginary observations.
The higher iss with respect to N, the stronger the effect of the prior on your posterior parameter estimates. With a fixed iss and a growing N, the Bayesian estimator and the maximum likelihood estimator converge to the same value.
A common rule of thumb is to use a small non-zero iss so that you avoid zero entries in the CPTs, corresponding to combinations that were not observed in the data. Such zero entries could then result in a network which generalizes poorly, such as some early versions of the Pathfinder system.
For more details on Bayesian parameter estimation you can have a look at the book by Koller and Friedman. I suppose many other Bayesian network books also cover the topic.
I would like to extract the alpha lagrange multipliers from the SVM function in the e1071 R package, however I am not sure if svm$coef is producing these?
Alphas are defined as in Equation 9.23, p352, An Introduction to Statistical Learning
In the documentation for SVM, it says that
SVM$Coefs = The corresponding coefficients times the training labels
Could someone please explain it?
$coefs produces alpha_i * y_i, but as alpha_i are by definition non-negative, you can simply take absolute value of coefs and it gives you Lagrange multipliers, and extract y_i by taking a sign (as they are only +1 or -1). This is just a simplification, often used in SVM packages, as multipliers are never actually used - only their product with the label, thus they are stored as a single number, for simplicity and efficiency, and in a case of need (like this one) - you can always reconstruct them.
First of all, I thank you all beforehand for reading this.
I am trying to fit a Standardized T-Student Distribution (i.e. a T-Student with standard deviation = 1) on a series of data; that is: I want to estimate the degrees of freedom via Maximum Likelihood Estimation.
An example of what I need to achieve can be found in the following (simple) Excel file I made:
https://www.dropbox.com/s/6wv6egzurxh4zap/Excel%20Implementation%20Example.xlsx?dl=0
Inside the Excel file, I have an image that contains the formula corresponding to the calculation of the loglikelihood function for the Standardized T Student Distribution. The formula was extracted from a Finance book (Elements of Financial Risk Management - by Peter Christoffersen).
So far, I have tried this with R:
copula.data <- read.csv(file.choose(),header = TRUE)
z1 <- copula.data[,1]
library(fitdistrplus)
ft1 = fitdist(z1, "t", method = "mle", start = 10)
df1=ft1$estimate[1]
df1
logLik(ft1)
df1 yields the number: 13.11855278779897
logLike(ft1) yields the number: -3600.2918050056487
However, the Excel file yields degrees of freedom of: 8.2962365022727, and a log-likelihood of: -3588.8879 (which is the right answer).
Note: the .csv file that my code reads is the following:
https://www.dropbox.com/s/nnh2jgq4fl6cm12/Data%20for%20T%20Copula.csv?dl=0
Any ideas? Thank you people!
The formula from your spreadsheet (with n, x substituted for the df parameter and the data)
=GAMMALN((n+1)/2)-GAMMALN(n/2)-LN(PI())/2-LN(n-2)/2-1/2*(1+n)*LN(1+x^2/(n-2))
or, exponentiating,
Gamma((n+1)/2) / (sqrt((n-2) pi) Gamma(n/2)) (1+x^2/(n-2))^-((n+1)/2)
?dt gives
f(x) = Gamma((n+1)/2) / (sqrt(n pi) Gamma(n/2)) (1 + x^2/n)^-((n+1)/2)
So the difference lies in those n-2 values in two places in the formula. I don't have enough context to see why the author is defining the t distribution in that different way; there may be some good reason ...
Looking at the negative log-likelihood curve directly, it certainly seems as though the fitdistrplus answer is agreeing with the direct calculation. (It would be very surprising if there were a bug in the dt() function, R's distribution functions are very broadly used and thoroughly tested.)
LL <- function(p,data=z1) {
-sum(dt(data,df=p,log=TRUE))
}
pvec <- seq(6,20,by=0.05)
Lvec <- sapply(pvec,LL)
par(las=1,bty="l")
plot(pvec,Lvec,type="l",
xlab="df parameter",ylab="negative log-likelihood")
## superimpose fitdistr results ...
abline(v=coef(ft1),lty=2)
abline(h=-logLik(ft1),lty=2)
Unless there's something else you're not telling us about the problem definition, it seems to me that R is getting the right answer. (The mean and sd of the data you gave were not exactly equal to 0 and 1 respectively, but they were close; centering and scaling gave an even larger value for the parameter.)
How to treat p value in R ?
I am expecting very low p values like:
1.00E-80
I need to -log10
-log10(1.00E-80)
-log10(0) is Inf, but Inf at sense of rounding too.
But is seems that after 1.00E-308, R yields 0.
1/10^308
[1] 1e-308
1/10^309
[1] 0
Is the accuracy of p-value display with lm function the same as the cutoff point, 1e-308, or it is just designed such that we need a cutoff point and I need to consider a different cutoff point - such as 1e-100 (for example) to replace 0 with <1e-100.
There are a variety of possible answers -- which one is most useful depends on the context:
R is indeed incapable under ordinary circumstances of storing floating-point values closer to zero than .Machine$double.xmin, which varies by platform but is typically (as you discovered) on the order of 1e-308. If you really need to work with numbers this small and can't find a way to work on the log scale directly, you need to search Stack Overflow or the R wiki for methods for dealing with arbitrary/extended precision values (but you probably should try to work on the log scale -- it will be much less of a hassle)
in many circumstances R actually computes p values on the (natural) log scale internally, and can if requested return the log values rather than exponentiating them before giving the answer. For example, dnorm(-100,log=TRUE) gives -5000.919. You can convert directly to the log10 scale (without exponentiating and then using log10) by dividing by log(10): dnorm(-100,log=TRUE)/log(10)=-2171, which would be too small to represent in floating point. For the p*** (cumulative distribution function) functions, use log.p=TRUE rather than log=TRUE. (This particular point depends heavily on your particular context. Even if you are not using built-in R functions you may be able to find a way to extract results on the log scale.)
in some cases R presents p-value results as being <2.2e-16 even when a more precise value is known: (t1 <- t.test(rnorm(10,100),rnorm(10,80)))
prints
....
t = 56.2902, df = 17.904, p-value < 2.2e-16
but you can still extract the precise p-value from the result
> t1$p.value
[1] 1.856174e-18
(in many cases this behaviour is controlled by the format.pval() function)
An illustration of how all this would work with lm:
d <- data.frame(x=rep(1:5,each=10))
set.seed(101)
d$y <- rnorm(50,mean=d$x,sd=0.0001)
lm1 <- lm(y~x,data=d)
summary(lm1) prints the p-value of the slope as <2.2e-16, but if we use coef(summary(lm1)) (which does not use the p-value formatting), we can see that the value is 9.690173e-203.
A more extreme case:
set.seed(101); d$y <- rnorm(50,mean=d$x,sd=1e-7)
lm2 <- lm(y~x,data=d)
coef(summary(lm2))
shows that the p-value has actually underflowed to zero. However, we can still get an answer on the log scale:
tval <- coef(summary(lm2))["x","t value"]
2*pt(abs(tval),df=48,lower.tail=FALSE,log.p=TRUE)/log(10)
gives -692.62 (you can check this approach with the previous example where the p-value doesn't overflow and see that you get the same answer as printed in the summary).
Small numbers are generally hard to deal with.
The limit in R for infinite is caused by the use of double precision floating point :
?double All R platforms are required to work with values conforming to the IEC 60559 (also known as IEEE 754) standard. This basically works with a precision of 53 bits, and represents to that precision a range of absolute values from about 2e-308 to 2e+308.
http://en.wikipedia.org/wiki/Double_precision_floating-point_format
You may find the Rmpfr package helpful here as it allows you to create multiple precision numbers.
install.packages("Rmpfr")
require(Rmpfr)
log(mpfr(1/10^309, precBits=500))