gzf200 at few.vu.nl
2009-Jun-27 12:15 UTC
[R] Regression; how to get t-values for all parameters estimates
Dear all, Even after a couple of hours looking at old messages I still haven't found a solution for my problem. I'm trying to fit an additive linear regression model with 2 effects, both fixed, to some dataset. The function contrasts(effectA) <- contr.sum can gaurantee that the coefficients per parameter sum to one, and the function dummy.coef provices the estimates of all coefficientss. But I would also like to be able to obtain the corresponding t-values for ALL parameters (not just the number of effects minus 1, provided by summary()). Does anyone know how to get (all of) them? Here comes what I've already tried: ## Try data:> Data <- rbinom(1000,50,.9); > Dates <- Sys.Date()-(1000:1) > facweek <- factor(weekdays(Dates,abbreviate=TRUE)) > facmonth <- factor(months(Dates,abbreviate=TRUE)) > contrasts(facweek) <- contr.sum ; > contrasts(facmonth) <- contr.sum ; > fit <- lm(formula = Data ~ facweek + facmonth) > summary(fit)Call: lm(formula = Data ~ facweek + facmonth) Residuals: Min 1Q Median 3Q Max -8.7498 -1.3774 0.1778 1.5108 5.1643 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 45.01452 0.06976 645.252 <2e-16 *** facweek1 -0.37114 0.16836 -2.204 0.0277 * facweek2 0.40360 0.16836 2.397 0.0167 * facweek3 -0.19918 0.16885 -1.180 0.2384 facweek4 -0.13689 0.16834 -0.813 0.4163 facweek5 -0.07049 0.16835 -0.419 0.6755 facweek6 0.40974 0.16836 2.434 0.0151 * facmonth1 -0.12046 0.22053 -0.546 0.5850 facmonth2 -0.10832 0.26155 -0.414 0.6789 facmonth3 0.25281 0.21731 1.163 0.2450 facmonth4 0.40161 0.22627 1.775 0.0762 . facmonth5 -0.37409 0.21731 -1.721 0.0855 . facmonth6 0.06645 0.26155 0.254 0.7995 facmonth7 0.13627 0.22509 0.605 0.5451 facmonth8 -0.04789 0.21731 -0.220 0.8256 facmonth9 -0.22910 0.21731 -1.054 0.2920 facmonth10 0.11752 0.22053 0.533 0.5942 facmonth11 -0.27233 0.21731 -1.253 0.2104 --- Signif. codes: 0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1 Residual standard error: 2.174 on 982 degrees of freedom Multiple R-squared: 0.02673, Adjusted R-squared: 0.009879 F-statistic: 1.586 on 17 and 982 DF, p-value: 0.06097> print(dummy.coef(fit),digits=3)Full coefficients are (Intercept): 45 facweek: Fri Mon Sat Sun Thu Tue -0.3711 0.4036 -0.1992 -0.1369 -0.0705 0.4097 facmonth: Apr Aug Dec Feb Jan Jul -0.1205 -0.1083 0.2528 0.4016 -0.3741 0.0664 (Intercept): facweek: Wed -0.0357 facmonth: Jun Mar May Nov Oct Sep 0.1363 -0.0479 -0.2291 0.1175 -0.2723 0.1775
Dieter Menne
2009-Jun-27 16:48 UTC
[R] Re gression; how to get t-values for all parameters estimates
gzf200 wrote:> > Even after a couple of hours looking at old messages I still haven't found > a > solution for my problem. > I'm trying to fit an additive linear regression model with 2 effects, both > fixed, to some dataset. The function contrasts(effectA) <- contr.sum can > gaurantee that the coefficients per parameter sum to one, and the function > dummy.coef provices the estimates of all coefficientss. But I would also > like to > be able to obtain the corresponding t-values for ALL parameters (not just > the > number of effects minus 1, provided by summary()). Does anyone know how to > get > (all of) them? > >Trying to fit months and weeks as a series of categorical variables and picking the asterisks is a path to disaster and hopefully not used to model the future economic of our economy. If you must, you can check function estimable in package gmodels, package multcomp, package contrast, and several functions in Hmisc. Dieter -- View this message in context: http://www.nabble.com/Regression--how-to-get-t-values-for-all-parameters-estimates-tp24232820p24234659.html Sent from the R help mailing list archive at Nabble.com.