[R] autoregressive spectral density estimate by andrews' plug-in method?
Hello! I would like to ask if there is in R a function that estimates the spectral density function of a stochastic series at frequency zero by the plug-in method, advocated by Andrews in his paper Heteroscedasticity and Autocorrelation Consistent Covariance Matrix Estimation, Econometrica, 59,817-858. I saw R has functions that employ Andrews' plug-in method using an AR(1) approximation for the estimation of the variance-covariance matrix in linear models. They come with the sandwich package. The so called meat is actually the estimate of the spectral density matrix of the model coefficients at frequency zero. However, I have a time series of length 160 and I need to estimate its spectral density via Andrews methodology. Any suggestions will be appreciated. Excuse me if I am asking something obvious. Regards, Martin - Нека музиката бъде с теб! http://musicidol.btv.bg/ __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] about the unscaled covariances from a summary.lm object
Hello! I want to clarify something about the unscaled covarinces component of a summary.lm object. So we have the regressor matrix X. If the fitted lm object is lmobj, the inverse of the matrix t(X)%*%X is xx, and the residual variance is sigma^2_e, the variance-covariance matrix of the OLS estimate of the coefficients is given by: xx*sigma^2_e I saw that what the function vcov actually does is simply: vcov=summary(lmobj)$sigma^2 * summary(lmobj)$cov.unscaled So the cov.unscaled component should give the matrix xx. I am right? I tried inverting the matrix t(X)%*%X with solve by issuing: solve(t(X)%*%X), but I get a matrix quite different from the matrix given by cov.unscaled. Is it just computational instability, or I am missing something important? Regards, Martin __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] modified test of phillips-perron
Hello! I would like to know whether the test of Phillips-Perron, modified for time series with large negative moving average terms implemented in R. Besides I would also like to ask the same question about the Leybourne-McCabe test for stationarity. Regards, Martin __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] about systemfit
Hello. I am still a newbie in R. Excuse me if I am asking something obvious. My efforts to get an answer through browsing the mailing archives failed. I want to perform an augmented Dickey-Fuller test and to obtain AIC and BIC and to be able to impose some linear restrictions on the ADF regression so as to decide the correct order of autoregression. However I could find no obvious way to impose linear restrictions or to obtain AIC from the the result of ADF.test from uroot. That is why I turned to systemfit. I ran the ADF regression with systemfit and obtained the same coefficient estimates as through ADF.test (as it had to be). Unfortunately I could not find how to extract AIC from the result of systemfit, so I evaluated the ADF regression by lm. So far so good. However the results of ftest.systemfit and linear.hypothesis from the car package are very different, while the results from waldtest.systemfit and linear.hypothesis coincide. I have no explanation for this issue a nd I could not see the code of linear.hypothesis. When I type linear.hypothesis I get: function (model, ...) UseMethod(linear.hypothesis) environment: namespace:car but when I type ftest.systemfit, I do see the actual code. Why? Anyway, here are the results in more detail: eqns-list(eq = y ~ trend+ x1 + x[,1] + x[,2] + x[,3] + x[,4] + x[,5] + x[,6] + x[,7] + x[,8] + x[,9] + x[,10] + x[,11] + x[,12] + x[,13] Rrestr10-matrix(0,10,16);Rrestr10[1,16]=Rrestr10[2,15]=Rrestr10[3,14]=Rrestr10[4,13]=Rrestr10[5,12]=Rrestr10[6,11]=Rrestr10[7,10]=Rrestr10[8,9]=Rrestr10[9,8]=Rrestr10[10,7]=1 adfResc-systemfit(method=OLS,eqns=eqns,R.restr=Rrestr10) adfResu-systemfit(method=OLS,eqns=eqns) adfResulm-lm(formula=eqns$eq) ftest.systemfit( object=adfResu, R.restr=Rrestr10) : F-test for linear parameter restrictions in equation systems F-statistic: 9.083 degrees of freedom of the numerator: 10 degrees of freedom of the denominator: 127 p-value: 3.449e-11 linear.hypothesis(model=adfResulm,hypothesis.matrix=Rrestr10,test=F): Res.Df RSS Df Sum of Sq F Pr(F) 1127 7.3782 2137 7.6848 -10 -0.3066 0.5277 0.868 As I said, the results of the chisquare test with linear.hypothesis and the waldtest.systemfit coincide. I have one more problem. This is the output of lrtest.systemfit: lrtest.systemfit(resultc=adfResc,resultu=adfResu) Likelihood-Ratio-test for parameter restrictions in equation systems LR-statistic: degrees of freedom: p-value: Why do I get empty values? In summary, I need to understand why the two ftests give different results; why lrtest.systemfit gives empty output; is there some way to extract AIC and BIC from object of class systemfit or from the result of ADF.test. Excuse me if I am asking something too obvious, but I am really at a loss. Any suggestions on any of the above questions will be welcomed. Regards, Martin __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] about systemfit
Thank you very much for your responsiveness. Here are the tests that show the same results, as they must: linear.hypothesis(model=adfResulm,hypothesis.matrix=Rrestr10,test=Chisq): Res.Df RSS Df Sum of Sq Chisq Pr(Chisq) 1127 7.3782 2137 7.6848 -10 -0.3066 5.2769 0.872 waldtest.systemfit(object=adfResu,R.restr=Rrestr10): Wald-test for linear parameter restrictions in equation systems Wald-statistic: 5.277 degrees of freedom: 10 p-value: 0.8719 So the Chisq test with linear.hypothesis from car and the waldtest.systemfit give the same output. But the F test with linear.hypothesis from car and ftest.systemfit give absolutely different results, as I demonstrated in my first post. I have no idea why. Regards, Martin __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] lag orders with ADF.test
Hello! I do not understand what is meant by: aic and bic follow a top-down strategy based on the Akaike's and Schwarz's information criteria in the datails to the ADF.test function. What does a top-down strategy mean? Probably the respective criterion is minimized and the mode vector contains the lag orders at which the criterion attains it local minima? When the calculation is over, the ADF.test function gives info about Lag orders. What are these lag orders? Are they the local minima of the criterion? I will be very thankful if you clarify this to me. I browsed a lot, but I could not find a clear answer. Thank you for your attention. Regards, Martin - http://auto-motor-und-sport.bg/ С бензин в кръвта! __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] adf test: trend, no drift - rep: invalid 'times' argument
Hello! I am applying the ADF.test function from package uroot to a time series of data. When I apply the full test, incorporating drift and trend terms, the regressor estimate of the drift term is not significantly different from zero. So I apply the test to a model without drift term, with deterministic trend only. But then I always get the following error: summary(ADF.test(wts=ts(seasons$summer, start=1850, frequency=1), itsd=c(0,1,c(0)), regvar=0, selectlags=list(mode=c(1,2,3 Error in rep(NA, ncol(table)) : invalid 'times' argument Error in summary(ADF.test(wts = ts(seasons$summer, start = 1850, frequency = 1), : error in evaluating the argument 'object' in selecting a method for function 'summary' I have no idea why this error occurs. Any suggestions will be appreciated. Regards, Martin - http://auto-motor-und-sport.bg/ С бензин в кръвта! __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] weighted MDS, alscal
Hello! I need to perform weighted multidimensional scaling analysis(WMDS). I did rsitesearch, googled, but I could find no info on how to perform WMDS using R. In several places they say it is possible with the ALSCAL algorithm, but I could not find the relevant function to carry it out. - Заложете на късмета си със Спортингбет! __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] analog to the matlab buffer function?
Hello! I am new to R. I could not find a function analogous to matlab's function buffer, which is used in signal processing. Is there such a function in R? What I need to do is as follows. If I apply the function to the vector c(1:5) for example with a window length 3 and overlapping 2, I need to get a matrix like this: 1 2 3 2 3 4 3 4 5 In matlab this is achieved with the function buffer. Is there ananalogous R function? Thank you very much in advance. Regards, Martin __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] nonlinear least squares fitting Trust-Region
Dear Mr Graves, Thank you very much for your response. Nobody else from this mailing list ventured to reply to me for the two weeks since I posted my question. nlminb and optim are just optimization procedures. What I need is not just optimization, but a nonlinear CURVE FITTING procedure. If there is some way to perform nonlinear curve fitting with the Trust-Region algorithm using any of these functions, I would me much obliged to you if you suggest to me how to achieve that. You asked me why I do not want Gauss-Newton. Since I am not an expert in the field of optimization, I am just conforming to what matlab documentation suggests, namely: Algorithm used for the fitting procedure: Trust-Region -- This is the default algorithm and must be used if you specify coefficient constraints. Levenberg-Marquardt -- If the trust-region algorithm does not produce a reasonable fit, and you do not have coefficient constraints, you should try the Levenberg-Marquardt algorithm. Gauss-Newton --THIS ALGORITHM IS POTENTIALLY FASTER THAN THE OTHER ALGORITHMS, BUT IT ASSUMES THAT THE RESIDUALS ARE CLOSE TO ZERO. IT IS INCLUDED FOR PEDAGOGICAL REASONS AND SHOULD BE THE LAST CHOICE FOR MOST MODELS AND DATA SETS. I browsed some literature about the garchfit function, but I did not see the Trust-Region algorithm there either: algorithm = c(sqp, nlminb, lbfgsb, nlminb+nm, lbfgsb+nm), control = list(), title = NULL, description = NULL, ...) Thank you for your attention. I am looking forward to your reply. Regards, Martin - vbox7.com - Забавни видео клипове! __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] nonlinear least squares trust region fitting ?
Hello! I am running R-2.3.1-i386-1 on Slackware Linux 10.2. I am a former matlab user, moving to R. In matlab, via the cftool, I performed nonlinear curve fitting using the method nonlinear least squares with the Trust-Region algorithm and not using robust fitting. Is it possible to perform the same analysis in R? I read quite a lot of R documentation, but I could not find an alternative solution. If there is such, please forgive my ignorance (I am a newbie in R) and tell me which function from which package is capable of performing the same analysis. If the same analysis is not possible to carry out in R, I would be grateful if you suggest to me some alternative procedure. I found that the nls function performs nonlinear least squares. The problem is that I do not want to implement the Gauss-Newton algorithm. In the worst case I would be contented with the Levenberg-Marquardt algorithm, if it is implemented in R. R nls's documentation mentions the port package and the ‘nl 2sol’ algorithm, but I could not find that package in the CRAN repository, so that I could read and judge whether that algorithm would be appropriate. Thank you very much in advance. I am looking forward to your answer. Regards, Martin - http://ide.li/ - портал за българите по света. Статии, новини, форуми, снимки, информация. __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] nonlinear least squares trust region fitting ?
Hello! I am running R-2.3.1-i386-1 on Slackware Linux 10.2. I am a former matlab user, moving to R. In matlab, via the cftool, I performed nonlinear curve fitting using the method nonlinear least squares with the Trust-Region algorithm and not using robust fitting. Is it possible to perform the same analysis in R? I read quite a lot of R documentation, but I could not find an alternative solution. If there is such, please forgive my ignorance (I am a newbie in R) and tell me which function from which package is capable of performing the same analysis. If the same analysis is not possible to carry out in R, I would be grateful if you suggest to me some alternative procedure. I found that the nls function performs nonlinear least squares. The problem is that I do not want to implement the Gauss-Newton algorithm. In the worst case I would be contented with the Levenberg-Marquardt algorithm, if it is implemented in R. R nls's documentation mentions the port package and the ‘nl 2sol’ algorithm, but I could not find that package in the CRAN repository, so that I could read and judge whether that algorithm would be appropriate. Thank you very much in advance. I am looking forward to your answer. Regards, Martin - http://ide.li/ - портал за българите по света. Статии, новини, форуми, снимки, информация. __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] NonLinearLeastSquares Trust-Region
Hello! I am running R-2.3.1-i386-1 on Slackware Linux 10.2. I am a former matlab user, moving to R. In matlab, via the cftool, I performed nonlinear curve fitting using the method nonlinear least squares with the Trust-Region algorithm and not using robust fitting. Is it possible to perform the same analysis in R? I read quite a lot of R documentation, but I could not find an alternative solution. If there is such, please forgive my ignorance (I am a newbie in R) and tell me which function from which package is capable of performing the same analysis. If the same analysis is not possible to carry out in R, I would be grateful if you suggest to me some alternative procedure. I found that the nls function performs nonlinear least squares. The problem is that I do not want to implement the Gauss-Newton algorithm. In the worst case I would be contented with the Levenberg-Marquardt algorithm, if it is implemented in R. R nls's documentation mentions the port package and the ‘nl2sol’ algorithm, but I could not find that package in the CRAN repository, so that I could read and judge whether that algorithm would be appropriate. Thank you very much in advance. I am looking forward to your answer. Regards, Martin - http://ide.li/ - портал за българите по света. Статии, новини, форуми, снимки, информация. __ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
