Re: [R] density vs. mass for discrete probability functions

Stefan--- Under the measure-theoretic approach to probability, discrete & continuous probability densities follow the same underlying mathematical principles. Check any text on measure-theoretic probability theory. ---JFL Stefan Schreiber Sent by: "R-help" 03/14/2019 08:43 PM To

Re: [R] Ghost variables

The script is complete. When I start, the environment is empty. The warnings are issued for these "ghost" variables well before they are created later in the script. Somehow previous incarnations are lingering around and being unhappy even after they were "deleted". Jeff Newmiller has

[R] Ghost variables

Fellow R-gonauts: I frequently erase/remove all the objects in my current environment so can I re-run scripts to ensure that analyses are complete, error-free, and accurate. However, sometimes when I re-rerun a script I get warning messages (see below for example) regarding some variables

[R] Unexpected YAML

R Users: Whenever I fire up R, I now get the following message (red) at the end of the prologue. R version 3.5.1 (2018-07-02) -- "Feather Spray" Copyright (C) 2018 The R Foundation for Statistical Computing Platform: x86_64-w64-mingw32/x64 (64-bit) R is free software and comes with ABSOLUTELY

Re: [R] Modèle à Equations structurelles

The more general model for estimating structural models is lavaan, contained in the package lavaan. yaya bamba via R-help Sent by: "R-help" 05/10/2018 12:01 PM Please respond to yaya bamba To "r-help@r-project.org"

Re: [R] Converting SAS Code

I wish to second this approach to learning R. I tried for several years to translate other stat programs or provide parallel analyses with R. This dabbling-in-R approach did not work . When a transferred to a research unit that could ill afford commercial software, I devoted my entire time

Re: [R] Sampe numbers

N <- 100 C <- 50 x <- numeric(N) for (i in 1:N){ x[i] <- sample(C-sum(x),1) } x sum(x) Frederic Ntirenganya Sent by: "R-help" 07/19/2016 01:41 PM To "r-help@r-project.org" , cc Subject [R] Sampe numbers Hi

Re: [R] Request for help

There is a video tutorial on the RStudio web site showing how to create R packages within RStudio. Hadley Wickham also has a book on creating R packages. Bert Gunter Sent by: "R-help" 06/03/2016 02:27 PM To suparna biswas

[R] promoting scalar arguments to vectors in a function

R users: Suppose I have a function that takes three numeric arguments x, y, and z, any of which may be scalars or vectors. Suppose further that the user takes one of the arguments, say y, as a vector with the other two as scalars. Is there an existing R function that will promote the other

Re: [R] Install.package CAIC

You should consider the package "caper", which provides a CAIC. From the caper manual "The caper package implements the methods originally provided in the programs CAIC (Purvis and Rambaut, 1995b) and MacroCAIC (Agapow and Isaac, 2002)." Hope this helps. Gabriela Wofkova

Re: [R] Correct notation for functions, packages when using LaTex

Erin There is a LaTeX package called listings. It has an R option to prettyprint code. Never used it. Joe "Kevin E. Thorpe" Sent by: "R-help" 12/10/2015 12:14 PM To Erin Hodgess , cc R help

Re: [R] Beta distribution approximate to Normal distribution

Scratch the rchisq (it should have been sqrt(rchisq), but that doesn't help.). Use the truncated normal u <- 3; a <- 2; N <- 100 x <- numeric(N) for (i in 1:N){ repeat{ if( (x[i] <- rnorm(1, u, a)) >= 0 ) break } } or the folded normal abs(rnorm(N, u, a)), They give similar

Re: [R] Beta distribution approximate to Normal distribution

Your question makes no sense as stated. However, guessing at what you want, you should perhaps consider the non-central chi-square density with 1 df and ncp = u/a, i.e, rchisq(100, df=1, ncp=u/a) Joe Joseph F. Lucke, PhD Senior Statistician Research Institute on Addictions University at

Re: [R] WEIBULL or EXPONENTIAL?

As a start you can use an exploratory approach. Standard survival analysis texts show you how to use a log-log plot to assess whether a distribution is Weibull. Of course, the exponential is a special case of the Weibull. CHIRIBOGA Xavier xavier.chirib...@unine.ch Sent by: R-help

Re: [R] generating phi using function()

Your function phi has 5 arguments with no defaults. Your call only has 3 arguments. Hence the error message. phi - function(w1, w2, j, k, K){ + zaehler - (k/K)^(w1-1)*(1-k/K)^(w2-1) + nenner - sum( ((1:K)/K)^(w1-1)*(1-(1:K)/K)^(w2-1)) + return( zaehler/nenner ) + } phi(c(1, 1), 44L,

Re: [R] Graph with ggplot2.

ylim(0%,100%) is not valild. It should be ylim(0,100). Jeff Newmiller jdnew...@dcn.davis.ca.us Sent by: R-help r-help-boun...@r-project.org 03/25/2015 11:14 AM To BenedettaB24 . benedetta.brune...@gmail.com, r-help@r-project.org r-help@r-project.org, cc Subject Re: [R] Graph with ggplot2.

Re: [R] Difference betweeen cor.test() and formula everyone says to use

The distribution of the statistic $ndf * r^2 / (1-r^2)$ with the true value $\rho = zero$ follows an $F(1,ndf)$ distribution. So the t-test is the correct test for $\rho=0$. Fisher's z is an asymptotically normal transformation for any value of $\rho$. Thus Fisher's z is better for testing

Re: [R] Inverse Student t-value

...@ria.buffalo.edu, cc Duncan Murdoch murdoch.dun...@gmail.com, r-help@r-project.org r-help@r-project.org, r-help-boun...@r-project.org Subject Re: [R] Inverse Student t-value Hi JLucke, Maybe the old excel function. TINV and T.INV.2T is same function for T wo-Tailed Inverse of the student`s t

Re: [R] Inverse Student t-value

My Excel (2013) returns exactly what R does. I used both T.INV and T.INV.T2There is no TINV. Has Excel been updated? Duncan Murdoch murdoch.dun...@gmail.com Sent by: r-help-boun...@r-project.org 09/30/2014 02:36 PM To Andre geomodel...@gmail.com, cc r-help@r-project.org Subject Re:

Re: [R] Inverse Student t-value

The website has your answer. The t-distribution is a regularized incomplete beta function. The incomplete beta function is given by R's pbeta function. You regularize it with R's beta function. Then you use R's uniroot function to find the inverse. Good homework problem. Andre

Re: [R] hypergeometric integral

#First, correct your code H - 0.7; t - 1; s - seq(0,t,0.1); x - 1-t/s; library(gsl) Gauss2F1 - function(H,x){ a - H-1/2; b - 1/2-H; c - H+1/2; # if(x=0 x1){ #hyperg_2F1(a,b,c,x) # }else{ hyperg_2F1(c-a,b,c,1-1/(1-x))/(1-x)^b # } } #Here #F - Gauss2F1(a,b,c,x) #and here

Re: [R] hypergeometric integral

I made a stupid mistake. My note at the end is irrelevant as is the code before the library statement. Instead, the upper bound to your integral must be around .5, not 1. integrate(Gauss2F1,lower=0,upper=.5,H=0.7) 0.4954698 with absolute error 5.5e-15

Re: [R] (no subject)

There's a comma missing after sep=). DBxy-read.csv(paste(C:/users/marta/desktop/DB,filename, sep=) sep=,,header=TRUE) should be DBxy-read.csv(paste(C:/users/marta/desktop/DB,filename, sep=), sep=,,header=TRUE) ^ marta valdes

Re: [R] gumbel distribution

The Weibull and Gumbel distributions are NOT the same. Frede Aakmann Tøgersen fr...@vestas.com Sent by: r-help-boun...@r-project.org 05/14/2014 02:01 AM To eliza botto eliza_bo...@hotmail.com, r-help@r-project.org r-help@r-project.org, cc Subject Re: [R] gumbel distribution Don't

Re: [R] R in Excel

Check out http://www.r-bloggers.com/a-million-ways-to-connect-r-and-excel/ for an overview. See section 4. Olga Albutova oda...@gmail.com Sent by: r-help-boun...@r-project.org 04/29/2014 11:05 AM To r-help@r-project.org, cc Subject [R] R in Excel Hello! I need some help. I'd

Re: [R] A vector of normal distributed values with a sum-to-zero constraint

The sum-to-zero constraint imposes a loss of one degree of freedom. Of N samples, only (N-1) can be random. Thus the solution is N - 100 x - rnorm(N-1) x - c(x, -sum(x)) sum(x) [1] -7.199102e-17 Boris Steipe boris.ste...@utoronto.ca Sent by: r-help-boun...@r-project.org

Re: [R] A vector of normal distributed values with a sum-to-zero constraint

Then what's wrong with centering your initial values around the mean? Marc Marí Dell'Olmo marceivi...@gmail.com 04/01/2014 10:56 AM To Boris Steipe boris.ste...@utoronto.ca, cc jlu...@ria.buffalo.edu, r-help@r-project.org r-help@r-project.org Subject Re: [R] A vector of normal distributed

Re: [R] Installing WinBUGS OpenBUGS on OS X 10.8.5

1. To install and run WBugs or Obugs on OSX you must use Parallels or other Windows emulator. 2. Another possibility is to install JAGS ( http://mcmc-jags.sourceforge.net/) via MacPorts (http://www.macports.org/ ) and use R2JAGS. Agony agony_...@yahoo.com Sent by:

Re: [R] psych package - Cronbach`s Alpha - warning message

I too don't understand the error message. However Coefficient Alpha is easily calculated. To wit: library(psych) tt - cbind(c(1,0.58) , c(0.58,1)) colnames(tt) = rownames(tt) = list(V1 , V2) p - nrow(tt) alpha - p/(p-1) * (1-sum(diag(tt))/sum(tt)) Numerous caveats regarding using C-Alpha

Re: [R] Adjusted R-squared formula in lm()

THE adjusted R^2 is [1-(1-R2)·(n-1)/(n-v-1)], which you call McNemar’s formula. It was actually proposed first by Fisher in 1924. Theil's formula is equal to Fisher's. Wherry's formula, as you give it, is correct but was proposed to estimate the cross-validated R2, which is different

Re: [R] Moore-Penrose Generalized determinant?

If the matrix is singular, the determinant of the matrix and its M-P inverse are both zero. Sean O'Riordain sean...@acm.org Sent by: r-help-boun...@r-project.org 03/14/2012 07:41 AM To r-help@r-project.org cc Subject [R] Moore-Penrose Generalized determinant? Is there a function in R

Re: [R] R versus R Studio output differences

First, be sure your R and the R associated with Rstudio are the same R versions. In Rstudio, check Tools - Options - R version. It looks as if your R-studio is running the 32-bit version of R. Aayush Raman ayushra...@gmail.com Sent by: r-help-boun...@r-project.org 03/09/2012 07:11 AM To

Re: [R] Bayesian functions for mle2 object

I would recommend using the new Bayesian package 'LaplacesDemon' available on CRAN. Ben Bolker bbol...@gmail.com Sent by: r-help-boun...@r-project.org 08/29/2011 02:50 PM To r-h...@stat.math.ethz.ch cc Subject Re: [R] Bayesian functions for mle2 object Billy.Requena billy.requena at

Re: [R] Using Mplus via R

Yes. If you want a program that does some of what Mplus does, use the lavaan package or the sem package. Dimitri Liakhovitski dimitri.liakhovit...@gmail.com Sent by: r-help-boun...@r-project.org 07/18/2011 05:36 PM To r-help r-help@r-project.org cc Subject [R] Using Mplus via R

Re: [R] Value of 'pi'

A transcendental number is not the zero of any polynomial with rational, not just integer, coefficients . Ravi Varadhan rvarad...@jhmi.edu Sent by: r-help-boun...@r-project.org 05/31/2011 10:12 AM To 'Bentley Coffey' bentleygcof...@gmail.com, Vincy Pyne vincy_p...@yahoo.ca cc

Re: [R] How to regress data into coefficients for a gamma function

Regression for the gamma distribution can be expressed as a generalized linear model. Check Chapter 8 of McCullagh, P. Nelder, J. A. (1989), Generalized linear models, Chapman Hall, London, UK. Walter Anderson wandrso...@gmail.com Sent by: r-help-boun...@r-project.org 03/29/2011 09:57

Re: [R] Strange R squared, possible error

lm(y~x+0) yields the regression on x without the constant, i.e., y=bx+e, not y = a +e derek jan.kac...@gmail.com Sent by: r-help-boun...@r-project.org 03/16/2011 03:49 PM To r-help@r-project.org cc Subject [R] Strange R squared, possible error k=lm(y~x) summary(k) returns R^2=0.9994

[R] Bug in lattice auto.key argument

The Lattice auto.key argument has a bug in R.12.2. R version 2.12.2 (2011-02-25) Platform: i386-pc-mingw32/i386 (32-bit) other attached packages: [1] lattice_0.19-17 loaded via a namespace (and not attached): [1] grid_2.12.2 If I set up my plot parameters as require(lattice)

Re: [R] A Math question

I just found out that the sum of 1+2+3+... is -1/12. !?! Crazy. Not really. Google on divergent series or Euler summation. The result explains why bosonic string theory must operate only in 26 dimensions! It would then appear that ... -3 - 2 - 1 + 0 +1 +2 +3 + ... is zero.

Re: [R] A Math question

Kjetil et al, Unlike finite sums, infinite sums are not commutative. To have commutativity, one must have absolute summability, that is, the sum of the absolute values of the terms must be finite. If one has absolute summability, the infinite sum exists and is unique. This sum is not

Re: [R] OT: Is randomization for targeted cancer therapies ethical?

Clearly inferior treatments are unethical. Donald Berry at MD Anderson in Houston TX and Jay Kadane at Carnegie Mellon have been working on more ethical designs within the Bayesian framework. In particular, response adaptive designs reduce the assignment to and continuation of patients on

Re: [R] confidence intervals around p-values

A confidence interval around the p-value makes no sense because there is no parameter being estimated, but the sampling distribution of the p-value makes a lot of sense. The pre-observational P-value is a random variable that is a function of the underlying random variable being tested. That

Re: [R] path analysis

There are three paths to path analysis in R: the SEM package; the LAVAAN package; and the OpenMx approach. The first two are R programs. The last accesses the program OpenMx. Guy rotem rottem...@gmail.com Sent by: r-help-boun...@r-project.org 09/06/2010 10:37 AM To r-help@r-project.org

Re: [R] Brown-Forsythe test of equality of MEANS

The following reference that contains a short Fortran program for the Brown-Forsythe ANOVA Reed, James F., I. Stark, D. B. Robust alternatives to traditional analyses of variance: Welch $W^*$, James $J_I^*$, James $J_II^*$, and Brown-Forsythe $BF^*$ Computer Methods and Programs in

Re: [R] Brown-Forsythe test of equality of MEANS

Yes. I too remember BMD, BMDP, punch cards, computer printouts, and 24-hour turn-around. However, I am also old enough that when someone asks for an antique method, I will gladly supply it if I can, if only for historical reasons. Bert Gunter gunter.ber...@gene.com 08/31/2010 04:41 PM To

Re: [R] Regression Error: Otherwise good variable causes singularity. Why?

There appears to be a problem in both regressions, as a singularity is also reported in the second regression analysis as well. It appears that the litrate variable is considered a factor in the first analysis and continuous in the second. There also appears to be collinearity between the

Re: [R] integral in R

help(integrate) Nathalie Gimenes ncgmsanc...@gmail.com Sent by: r-help-boun...@r-project.org 07/19/2010 04:59 PM To r-help@r-project.org cc Subject [R] integral in R Hello All, I have to create a variable that is a function of another one (already created), its cumulative

Re: [R] Popularity of R, SAS, SPSS, Stata...

Yes. When my young son told someone that I was a doctor, he had to further explain I was a number doctor, not a people doctor. Barry Rowlingson b.rowling...@lancaster.ac.uk Sent by: r-help-boun...@r-project.org 06/22/2010 09:45 AM To Marc Schwartz marc_schwa...@me.com cc

Re: [R] What does LOESS stand for?

The closest explanation of the term loess I can find is in Local regression models by Cleveland, W. S.; Grosse, E. Shyu, W. M. (Chapter 8 of Statistical models in S (1992) by Chambers, J. M. Hastie, T. J.) ... local regression models is called loess, which is short for local regression,

Re: [R] Expectation of E(x^1/2)

Ben, Ravi, Chuck, and Haneef, Note that the standard normal density can be written (in pseudo-TeX) as 1/sqrt(2*pi) * exp[- ( x/sqrt(2) )^2]. The exponential on the right is a special case of the stretched exponential function exp[- (x/sqrt(tau))^beta]. The stretched exponential function has

Re: [R] Cross-validation for parameter selection (glm/logit)

Jay Unless I have misunderstood some statistical subtleties, you can use the AIC in place of actual cross-validation, as the AIC is asymptotically equivalent to leave-out-one cross-validation under MLE. Joe Stone, M. An asymptotic equivalence of choice of model by cross-validation and Akaike's

[R] Stacking matrices

What is an easy way to stack a matrix multiple times? E.g. I have a 6x6 matrix that I need to stack vertically 154 times to get a 6*154 by 6 matrix. I would rather not rbind(X,X,...,X) matrices.--Joe [[alternative HTML version deleted]]

Re: [R] OT Sorta: Odds Are, It's Wrong...

Which in turn reminds me of Box Tiao's line Since small differences in probability cannot be appreciated by the human mind, there seems little point in being excessively precise about uncertainty. Box, G. E. P. Tiao, G. C. (1973), Bayesian inference in statistical analysis, Addison-Wesley,

Re: [R] about IRT simulation

Helen Your program makes no sense. Try the following script I - 10 J - 5 response - matrix(0, 10, 5) #function for probability of response #revised but equivalent version of Helen's formula pij - function(a,b,theta) {1/(1+exp(-a*(theta-b)))} a - rnorm(J, 0.8, 0.04) a b - rnorm(J, 0, 1) b

Re: [R] code

?density ?max Rahim Alhamzawi rahimalhamz...@yahoo.co.uk Sent by: r-help-boun...@r-project.org 12/09/2009 05:23 AM To r-h...@stat.math.ethz.ch cc Subject [R] code Dear support, I want to compute the highest probability density for any data. please do you have any code to help me in

Re: [R] lm: RME vs. ML

You need to give your criteria for preferable. For normal-linear models, REML estimates of variances are unbiased, whereas ML estimates are downwardly biased. My intuition is that the ML-induced bias would be worse in small samples. I don't know about other distributions. Likewise I don't

Re: [R] The equivalence of t.test and the hypothesis testing of one way ANOVA

Guido wrote However, using a transformation matrix one can transform a model assuming unequal variances into an equivalent model assuming equal variances. On such a transformed model the F test or T test can be applied. This is indeed news to me. I thought such transformations for unequal

Re: [R] Bayesian regression stepwise function?

The BIC (Raftery) can be used for quasi-Bayesian model selection, but it's not stepwise. Ntzoufras shows how to use WinBUGS to conduct Bayesian model selection, but again it's not stepwise Ntzoufras, I. (2002), 'Gibbs variable selection using BUGS', Journal of Statistical Software 7(7),

Re: [R] Parameters of Beta distribution

Rescale your data x to (x-A)/(B-A). Maithili Shiva maithili_sh...@yahoo.com Sent by: r-help-boun...@r-project.org 10/07/2009 08:39 AM To r-help@r-project.org cc Subject [R] Parameters of Beta distribution Supose I have a data pertaining to credit loss as amounts -

[R] Unable to load rjags

The downloaded packages are in C:\Documents and Settings\jlucke\Local Settings\Temp\RtmpRAkMZy\downloaded_packages updating HTML package descriptions BUT, library(rjags) Error in fun(...) : Failed to locate JAGS 1.0.3 installation Error : .onLoad failed in 'loadNamespace' for 'rjags' Error: package

Re: [R] Logistic Regression for Multinomial Data using R

Use polr from the MASS package Nimal Fernando pnp...@gmail.com Sent by: r-help-boun...@r-project.org 09/25/2009 12:33 PM To r-help@r-project.org cc Subject Re: [R] Logistic Regression for Multinomial Data using R Hi I want to do logistic regression for multinomial data. How can I do

Re: [R] P-value and R-squared variable selection criteria

Lucas This problem is very old --- older than keypunches. There are several methods for selecting variables (forward, backwards, both, all subsets) using a variety of criteria (p-values, R^2, adjusted R^2, Cp, AIC, BIC, and more). Be sure you understand the methods, especially the tendency to

Re: [R] Contrasts within ANOVA frame (Repost)

Dr Shen Even though you pose 3 contrasts, there are only 2 degrees of freedom as the 3rd contrast is the sum of the first two. Your aov output states this as StdLot has only 2 df. Having used up your df, the aov has nothing to report. Joe Joseph F. Lucke Senior Statistician Research

Re: [R] integrate lgamma from 0 to Inf

I am late to this discussion so forgive me if I am being redundant. It appears to me that the integral from 0 to infinity of log-gamma may diverge to infinity. Equation 6.1.50 of Abramowitz Stegun shows that a re-expression of lnGamma(x) and the Stirling approximation involves

Re: [R] large factorials

I am late to this discussion so I apologize if what I say has already been noted. The function below does NOT require large factorials, but very small reciprocal of factorials. Thus if v2 would be replaced with something like exp(-lgamma(i-n+m+1)-lgamma(m+1)) and checked for underflow (ie

Re: [R] large factorials

The modified function I presented contains a stupid error, which is corrected below. sum1 - function(l,u,t,i,n,w,tol=.Machine$double.xmin) { v - 0 v2 - 1 for (m in 0 :w v2 tol) { v1 - ((u^(1/2))*(l^(1/2))*t)^(i-n+2*m) v2 - exp(-lgamma(i-n+m+1)-lgamma(m+1)) v3 - v1*v2 v - v+v3 } return(v)

Re: [R] integrate lgamma from 0 to Inf

IMHO, you should consult an advanced text on calculus with integration over infinite intervals, so that you understand what you are trying to do. Joseph F. Lucke Senior Statistician Research Institute on Addictions University at Buffalo State University of New York 1021 Main Street Buffalo, NY

Re: [R] Maple and R

You can export the Maple code to C or Fortran. I actually found it easier to export it to Fortran and then use a text editor to change Fortran's assignment ='s to R's -. After additional tweaks for R, you can convert the script into an R function. Joseph F. Lucke Senior Statistician

Re: [R] Maple and R

I forgot. You can get the Bessel functions from Robin Hankin's GSL package. R. K. S. Hankin 2006. Introducing gsl, a wrapper for the Gnu Scientific Library. Rnews 6(4):24-26 Joseph F. Lucke Senior Statistician Research Institute on Addictions University at Buffalo State University of New

[R] Distinguishing variables from functions with the same name

guRus: I have a variable beta as an argument to R's beta function. So essentially I have a case of beta(alpha, beta). What surprises me is that R doesn't barf on this stupid programming practice. R gets the right answer. How does R know beta the variable from beta the function? Joseph F.

Re: [R] get top 50 correlated item from a correlation matrix for each item

A solution using a toy example r - cor(mvrnorm(1000,mu=rep(0,10),Sigma=diag(10))) #assume a 10 x 10 matrix j - i-1:dim(r)[1] #generate matrix indices lt - outer(i,j,'') #get boolean lower triangle sort(r[lt],decreasing=TRUE)[1:5] #extract top 5 correlations Joseph F. Lucke Senior

[R] Can't install rggobi package

I am unable to install the rggobi package. I do not have trouble installing other packages. Any advice or suggestions? sessionInfo() R version 2.8.1 (2008-12-22) i386-pc-mingw32 locale: LC_COLLATE=English_United States.1252;LC_CTYPE=English_United States.1252;LC_MONETARY=English_United

[R] OT: Reference for SIC/BIC

Generally smart people, I need a recent reference for using the Schwarz/Bayesian information criterion for model fitting in a frequentist stepwise regression. Joe [[alternative HTML version deleted]] __ R-help@r-project.org mailing list

Re: [R] OT: Adding verbatim R code text into LaTeX documents: texttt; verb or url?

LaTeX offers a verbatim environment. \begin{verbatim} This is maintained verbatim, Latex commands and environments are typeset as written without any processing. \end{verbatim} Be sure to use the package verbatim. ---Joe Peter Dunn pdu...@usc.edu.au Sent by: r-help-boun...@r-project.org

Re: [R] using Sweave with a master file that has several iputted .tex files

Use the include function in LaTeX. Christopher W. Ryan cr...@binghamton.edu Sent by: r-help-boun...@r-project.org 01/27/2009 12:48 PM To r-h...@stat.math.ethz.ch cc Subject [R] using Sweave with a master file that has several iputted .tex files Suppose I have a Master.Rnw file that

[R] Cannot access packages

Although I have installed R many times on many machines without problems, I have recently encountered a problem with accessing packages. R has been installed successfully with the --internet2 option. The session info is below. sessionInfo() R version 2.8.1 (2008-12-22) i386-pc-mingw32