Zelig July 2006

zelig@lists.gking.harvard.edu

7 participants
8 discussions

Start a nNew thread

[RE]Re: [zelig] Obtaining probabilities in oprobit

by dysgraphia2325

17 years, 9 months

multinomial logit with zelig - baseline problem

by Joerg Mueller-Scheessel

Hello all, I tried to estimate a multinomial logit using zelig, mlogit. My problem is that I do not know how to determine the base variable. I red in the mailing archive from last year, that in the case of one formula the baseline is the last for numeric endogenous variables. But the case of different formula for each level is not mentioned so I expected that the number, which is not included in the formulas is the baseline. So I tried the examples on http://gking.harvard.edu/zelig/docs/Examples21.html <http://gking.harvard.edu/zelig/docs/Examples21.html> and changed in the case of different formula for each level the levels, but the results didn't change. Here is the "unchanged" example output: data(mexico) > z.out2 <- zelig(list(id(vote88, "1") ~ pristr + othcok, + id(vote88, "2") ~ othsocok), + model = "mlogit", data = mexico) > summary(z.out2) Call: zelig(formula = list(id(vote88, "1") ~ pristr + othcok, id(vote88, "2") ~ othsocok), model = "mlogit", data = mexico) Pearson Residuals: Min 1Q Median 3Q Max log(mu[,1]/mu[,3]) -4.1 -0.72 0.29 0.738 2.0 log(mu[,2]/mu[,3]) -2.1 -0.47 -0.21 -0.086 4.6 Coefficients: Value Std. Error t value (Intercept):1 2.20 0.276 8.0 (Intercept):2 -0.40 0.218 -1.9 pristr 0.66 0.077 8.6 othcok -1.19 0.091 -13.2 othsocok 0.21 0.138 1.5 Number of linear predictors: 2 Names of linear predictors: log(mu[,1]/mu[,3]), log(mu[,2]/mu[,3]) Dispersion Parameter for multinomial family: 1 Residual Deviance: 2367 on 2713 degrees of freedom Log-likelihood: -1183 on 2713 degrees of freedom Number of Iterations: 4 And for this run I have substituted in the second formula the 2 by a 3. But the output didn't change. z.out2 <- zelig(list(id(vote88, "1") ~ pristr + othcok, + id(vote88, "3") ~ othsocok), + model = "mlogit", data = mexico) > summary(z.out2) Call: zelig(formula = list(id(vote88, "1") ~ pristr + othcok, id(vote88, "3") ~ othsocok), model = "mlogit", data = mexico) Pearson Residuals: Min 1Q Median 3Q Max log(mu[,1]/mu[,3]) -4.1 -0.72 0.29 0.738 2.0 log(mu[,2]/mu[,3]) -2.1 -0.47 -0.21 -0.086 4.6 Coefficients: Value Std. Error t value (Intercept):1 2.20 0.276 8.0 (Intercept):2 -0.40 0.218 -1.9 pristr 0.66 0.077 8.6 othcok -1.19 0.091 -13.2 othsocok 0.21 0.138 1.5 Number of linear predictors: 2 Names of linear predictors: log(mu[,1]/mu[,3]), log(mu[,2]/mu[,3]) Dispersion Parameter for multinomial family: 1 Residual Deviance: 2367 on 2713 degrees of freedom Log-likelihood: -1183 on 2713 degrees of freedom Number of Iterations: 4 So, what have I done wrong? Thank you in advance! Joerg Mueller-Scheessel

17 years, 9 months

bug in setx?

by Eduardo Leoni

Hi - when I use setx whith a factor that has extra levels not in the dataset it gives "non-conformable arguments". dta <- data.frame(y=rnorm(10),x1=1:10,x2=factor(c(rep(1,3),rep(2,4),rep(3,3)))) z1 <- zelig(y~x1+x2,model="ls",data=dta) v1 <- setx(z1) s1 <- sim(z1,x=v1) ##this works ##now change data, coding factor 3 as 2 dta$x2[7:10] <- 2 z1 <- zelig(y~x1+x2,model="ls",data=dta) v1 <- setx(z1) Warning message: There is more than one mode. The first level is selected. in: FUN(X[[1]], ...) > s1 <- sim(z1,x=v1) Error in coef %*% t(x) : non-conformable arguments If this is indeed a bug, a possible solution is to set somewhere in setx function for (i in 1:ncol(dta)) { vnow <- dta[,i] dta[,i] <- vnow[drop=TRUE] } but may have unintended consequences that I am not aware of. best, -eduardo - Zelig Mailing List, served by Harvard-MIT Data Center Send messages: zelig(a)lists.gking.harvard.edu [un]subscribe Options: http://lists.gking.harvard.edu/?info=zelig Zelig program information: http://gking.harvard.edu/zelig/

17 years, 9 months

Obtaining probabilities in oprobit

by dysgraphia2325

17 years, 9 months

Re: zelig question

by Kosuke Imai

Hi Alexis, I think most models should be able to incoporate the weights by passing "weights" as an optional argument in zelig(). Kosuke On Mon, 17 Jul 2006, Alexis Diamond wrote: > hi kosuke, > > hope you are having a wonderful day! :) > > i was wondering if zelig has the capability to accept observation-specific > weights. didn't see it in the documentation. did i miss it? jens and i are > working on a consulting project for the government of germany, and we would > like to propose that they use Zelig (very intensively), but the thing is, > their dataset involves observation-specific weights. any suggestions? > > thank you, > > alexis > - Zelig Mailing List, served by Harvard-MIT Data Center Send messages: zelig(a)lists.gking.harvard.edu [un]subscribe Options: http://lists.gking.harvard.edu/?info=zelig Zelig program information: http://gking.harvard.edu/zelig/

17 years, 9 months

Re: [zelig] Ordered Probit in Zelig/R and EasyReg

by dysgraphia2325＠lycos.com

Thanks for the tip Olivia! Cheers, Peter > ----- Original Message ----- > From: "Olivia Lau" <olau(a)fas.harvard.edu> > To: "Gary King" <king(a)harvard.edu> > Subject: Re: [zelig] Ordered Probit in Zelig/R and EasyReg > Date: Sat, 15 Jul 2006 16:36:30 -0400 > > > Sidebar: As a general rule, one should never ever attach(data) when data is > an argument to a function that takes a formula. You will confuse R. Best, > Olivia > > On 7/15/06, Gary King <king(a)harvard.edu> wrote: > > > > > > you don't need to import results from Zelig into excel to compute > > quantities of interest, since Zelig will do it for you automatically. See > > the setx() and sim() steps. > > Gary King > > > > On Sat, 15 Jul 2006, dysgraphia2325(a)lycos.com wrote: > > > > > Using Win XP, current R and Zelig. > > > I am new to R and Zelig. (tail snipped for brevity) -- _______________________________________________ Search for businesses by name, location, or phone number. -Lycos Yellow Pages http://r.lycos.com/r/yp_emailfooter/http://yellowpages.lycos.com/default.as… - Zelig Mailing List, served by Harvard-MIT Data Center Send messages: zelig(a)lists.gking.harvard.edu [un]subscribe Options: http://lists.gking.harvard.edu/?info=zelig Zelig program information: http://gking.harvard.edu/zelig/

17 years, 9 months

Ordered Probit in Zelig/R and EasyReg

by dysgraphia2325＠lycos.com

Using Win XP, current R and Zelig. I am new to R and Zelig. The ability to program Zelig/R is valuable....not available in EasyReg In order to familiarize myself with the programs I ran the ordered probit model in Zelig then in EasyReg using the same dataset. Here are output details for my R session: > myFile = read.table(file = "C:/TestSetForR.txt",header=T) > attach(myFile) > z.out <- zelig(as.factor(Grade_1_8) ~ X1 + X2 + X3 + X4, model = "oprobit", data = myFile) > summary(z.out) Call: zelig(formula = as.factor(Grade_1_8) ~ X1 + X2 + X3 + X4, model = "oprobit", data = myFile) Coefficients: Value Std. Error t value X1 -0.054702 0.018865 -2.900 X2 0.005183 0.001054 4.919 X3 0.085056 0.016026 5.307 X4 -0.088101 0.014033 -6.278 Intercepts: Value Std. Error t value 1|2 -4.857 1.006 -4.830 2|3 -4.383 1.003 -4.370 3|4 -4.055 1.002 -4.048 4|5 -3.775 1.001 -3.770 5|6 -3.519 1.001 -3.516 6|7 -3.093 1.000 -3.093 7|8 -2.516 1.000 -2.517 Residual Deviance: 3843.73 AIC: 3865.73 ################################################ Here are output details for the same data run in EasyReg. EasyReg International [April 2, 2006] Session date: Friday July 14, 2006 Session time: 07:48:11 ---------------------------------------------- Ordered Probit model: Dependent variable: Y = Grade_1_8 Characteristics: Grade_1_8 First observation = 1 Last observation = 1009 Number of usable observations: 1009 Minimum value: 1.0000000E+000 Maximum value: 8.0000000E+000 Sample mean: 5.2893954E+000 This variable is integer valued. A discrete dependent variable model is suitable. X variables: X(1) = X1 X(2) = X2 X(3) = X3 X(4) = X4 X(5) = 1 Model: P(Y-1=0|x) = F(-b'x) P(Y-1=1|x) = F(-b'x+exp(b(6)))- F(-b'x) For j=2,..,m-1, P(Y-1=j|x) = F(-b'x+exp(b(6))+..+exp(b(5+j))) - F(-b'x+exp(b(6))+..+exp(b(4+j))) and P(Y-1=m|x) = 1 - F(-b'x+exp(b(6))+..+exp(b(4+m))) where m =7, b'x = b(1)x(1)+..+b(5)x(5), and F(u) = c.d.f. of N(0,1) distribution. Newton iteration succesfully completed after 30 iterations Last absolute parameter change = 0.0001 Last percentage change of the likelihood = -0.0459 Maximum likelihood estimation results: Variable Par. ML estimate t-value [p-value] x(1)=X1 b(1) -0.0547016 -2.87 [0.00407] x(2)=X2 b(2) 0.0051827 4.75 [0.00000] x(3)=X3 b(3) 0.0850551 5.29 [0.00000] x(4)=X4 b(4) -0.0881011 -6.25 [0.00000] x(5)=1 b(5) 4.8565051 4.74 [0.00000] b(6) -0.7482279 -7.65 [0.00000] b(7) -1.1145713 -10.92 [0.00000] b(8) -1.2712482 -12.77 [0.00000] b(9) -1.3635656 -13.48 [0.00000] b(10) -0.8530407 -11.26 [0.00000] b(11) -0.5500365 -8.14 [0.00000] [The two-sided p-values are based on the normal approximation] Log likelihood: -1.05015155949E+003 Sample size (n): 1009 Information criteria: Akaike: 2.103373 Hannan-Quinn: 2.123736 Schwarz: 2.156974 In the main the independent coefficients and t-values were quite close in value except that the intercepts are dissimilar...not unexpected as the contrasts are defined differently, as differences. As shown above, EasyReg provides expressions for the probabilities for the various levels viz P(Y-1=0|x) = F(-b'x) P(Y-1=1|x) = F(-b'x+exp(b(6)))- F(-b'x) For j=2,..,m-1, P(Y-1=j|x) = F(-b'x+exp(b(6))+..+exp(b(5+j))) - F(-b'x+exp(b(6))+..+exp(b(4+j))) and P(Y-1=m|x) = 1 - F(-b'x+exp(b(6))+..+exp(b(4+m))) where m =7, b'x = b(1)x(1)+..+b(5)x(5), and F(u) = c.d.f. of N(0,1) distribution. This is useful for me as I enter these equations into an Excel spreadsheet, useful to test holdout samples and for teaching purposes for example. My question: in the Zelig documentation does there exist equivalent equations to calculate probabilities of the different levels? Any advice appreciated!....Cheers, Peter. -- _______________________________________________ Search for businesses by name, location, or phone number. -Lycos Yellow Pages http://r.lycos.com/r/yp_emailfooter/http://yellowpages.lycos.com/default.as… - Zelig Mailing List, served by Harvard-MIT Data Center Send messages: zelig(a)lists.gking.harvard.edu [un]subscribe Options: http://lists.gking.harvard.edu/?info=zelig Zelig program information: http://gking.harvard.edu/zelig/

17 years, 9 months

Sandwich estimator

by Brian Perron

The Zelig package is fantastic. I am still new to R and have found the functionality of Zelig truly impressive. Two quick questions: 1) Is there a way to invoke a sandwich estimator for clustered observations, similar to the robcov function in Harrell's Hmisc package? If not, could I change the standard errors manually in from a model (e.g., logit) generated via zelig? I am guessing there is a way, since just about anything is possible with R. (If somebody could provide code for doing this, that would be great.) 2) Are there any plans for adding Cox regression to the available stock of survival models? Thanks in advance, Brian - Zelig Mailing List, served by Harvard-MIT Data Center Send messages: zelig(a)lists.gking.harvard.edu [un]subscribe Options: http://lists.gking.harvard.edu/?info=zelig Zelig program information: http://gking.harvard.edu/zelig/

17 years, 9 months

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

Zelig July 2006