Hello, 1) I am interested in performing a limited-dependent variable linear regression. By this I mean a classical linear regression, but for the case where the values of the dependent variable cannot vary from -infinity to +infinity, but are truncated and so are between two finite limits L1 and L2. Does R1.7 have this capability? If so what is (are) the relevant command(s)? 2) I am also interested in sampling from a multivariate normal distribution. I see that R has a command mvrnorm, but when I try to use it, I get a message to the effect that the command is not recognized. Is there something I must do to enable this command, or is it not available in version 1.7? Thanks, Howard [[alternate HTML version deleted]]
Regarding the second question, did you try "library(mass)" before your mvrnorm command? Regarding the first question: I'm not certain what you want. The function "lm", for example, minimizes the sums of squares of deviations between observed and predicted. This also maximizes the likelihood assuming the deviations between observed and the "true model" are independent normal variates with constant variance. Can you express your model in terms of a probability density function for your observations with the model related in some way to parameters of that probability density? hope this helps. spencer graves Howard Alper wrote:> > Hello, > > 1) I am interested in performing a limited-dependent variable linear regression. By this I mean a classical linear regression, but for the case where the values of the dependent variable cannot vary from -infinity to +infinity, but are truncated and so are between two finite limits L1 and L2. Does R1.7 have this capability? If so what is (are) the relevant command(s)? > > 2) I am also interested in sampling from a multivariate normal distribution. I see that R has a command mvrnorm, but when I try to use it, I get a message to the effect that the command is not recognized. Is there something I must do to enable this command, or is it not available in version 1.7? > > Thanks, > > Howard > > > > [[alternate HTML version deleted]] > > ______________________________________________ > R-help at stat.math.ethz.ch mailing list > https://www.stat.math.ethz.ch/mailman/listinfo/r-help
On Tue, 06 May 2003 15:27:20 -0400, you wrote in message <seb7d46c.048 at healthsmtp.nycnet>:> > Hello, > >1) I am interested in performing a limited-dependent variable linear regression. By this I mean a classical linear regression, but for the case where the values of the dependent variable cannot vary from -infinity to +infinity, but are truncated and so are between two finite limits L1 and L2. Does R1.7 have this capability? If so what is (are) the relevant command(s)?I don't think that's built in, but it's always worth looking at the contributed packages on cran.r-project.org just in case. There are general purpose likelihood maximizers (nlm, optim).>2) I am also interested in sampling from a multivariate normal distribution. I see that R has a command mvrnorm, but when I try to use it, I get a message to the effect that the command is not recognized. Is there something I must do to enable this command, or is it not available in version 1.7?When the help says "mvrnorm(MASS)", that means that the function is in the package called "MASS". You need to execute library(MASS) to get that function loaded onto the search path. You can see which packages are already there by executing search() Duncan Murdoch
Duncan Murdoch wrote:> > On Tue, 06 May 2003 15:27:20 -0400, you wrote in message > <seb7d46c.048 at healthsmtp.nycnet>: > ... > > >2) I am also interested in sampling from a multivariate normal distribution. I see that R has a command mvrnorm, but when I try to use it, I get a message to the effect that the command is not recognized. Is there something I must do to enable this command, or is it not available in version 1.7? > > When the help says "mvrnorm(MASS)", that means that the function is in > the package called "MASS". You need to execute > > library(MASS) > > to get that function loaded onto the search path. You can see which > packages are already there by executing > > search() > > Duncan MurdochYou can also sample a multivariate normal distribution using rmvnorm from a contributed package "mvtnorm". Execute library(mvtnorm) before using it. Paul Peng
You can use the optim() function in MASS to optimize a loglikelihood function. For inspiration on the form of the loglikelihoodfunction in your case, you may consult G.S. Maddala "Limited dependent variables in econometrics" 1983 (the title may not be completely correct). Ruud *********** REPLY SEPARATOR *********** On 5/6/2003 at 3:27 PM Howard Alper wrote:>Hello, > >1) I am interested in performing a limited-dependent variable linear >regression. By this I mean a classical linear regression, but for the >case where the values of the dependent variable cannot vary from -infinity >to +infinity, but are truncated and so are between two finite limits L1 >and L2. Does R1.7 have this capability? If so what is (are) the relevant >command(s)? > >2) I am also interested in sampling from a multivariate normal >distribution. I see that R has a command mvrnorm, but when I try to use >it, I get a message to the effect that the command is not recognized. Is >there something I must do to enable this command, or is it not available >in version 1.7? > > Thanks, > > Howard > > > > [[alternate HTML version deleted]] > >______________________________________________ >R-help at stat.math.ethz.ch mailing list >https://www.stat.math.ethz.ch/mailman/listinfo/r-help