Estimates the location and scale parameters of the logistic distribution by maximum likelihood estimation.

logistic1(llocation = "identitylink", scale.arg = 1, imethod = 1)
logistic(llocation = "identitylink", lscale = "loglink",
         ilocation = NULL, iscale = NULL, imethod = 1, zero = "scale")

Arguments

llocation, lscale

Parameter link functions applied to the location parameter \(l\) and scale parameter \(s\). See Links for more choices, and CommonVGAMffArguments for more information.

scale.arg

Known positive scale parameter (called \(s\) below).

ilocation, iscale

See CommonVGAMffArguments for information.

imethod, zero

See CommonVGAMffArguments for information.

Details

The two-parameter logistic distribution has a density that can be written as $$f(y;l,s) = \frac{\exp[-(y-l)/s]}{ s\left( 1 + \exp[-(y-l)/s] \right)^2}$$ where \(s > 0\) is the scale parameter, and \(l\) is the location parameter. The response \(-\infty<y<\infty\). The mean of \(Y\) (which is the fitted value) is \(l\) and its variance is \(\pi^2 s^2 / 3\).

A logistic distribution with scale = 0.65 (see dlogis) resembles dt with df = 7; see logistic1 and studentt.

logistic1 estimates the location parameter only while logistic estimates both parameters. By default, \(\eta_1 = l\) and \(\eta_2 = \log(s)\) for logistic.

logistic can handle multiple responses.

Value

An object of class "vglmff" (see vglmff-class). The object is used by modelling functions such as vglm, rrvglm and vgam.

References

Johnson, N. L. and Kotz, S. and Balakrishnan, N. (1994). Continuous Univariate Distributions, 2nd edition, Volume 1, New York: Wiley. Chapter 15.

Forbes, C., Evans, M., Hastings, N. and Peacock, B. (2011). Statistical Distributions, Hoboken, NJ, USA: John Wiley and Sons, Fourth edition.

Castillo, E., Hadi, A. S., Balakrishnan, N. and Sarabia, J. S. (2005). Extreme Value and Related Models with Applications in Engineering and Science, Hoboken, NJ, USA: Wiley-Interscience, p.130.

deCani, J. S. and Stine, R. A. (1986). A Note on Deriving the Information Matrix for a Logistic Distribution, The American Statistician, 40, 220–222.

Author

T. W. Yee

Note

Fisher scoring is used, and the Fisher information matrix is diagonal.

Examples

# Location unknown, scale known
ldata <- data.frame(x2 = runif(nn <- 500))
ldata <- transform(ldata, y1 = rlogis(nn, loc = 1+5*x2, sc = exp(2)))
fit1 <- vglm(y1 ~ x2, logistic1(scale = exp(2)), ldata, trace = TRUE)
#> Iteration 1: loglikelihood = -1980.8368
#> Iteration 2: loglikelihood = -1980.7961
#> Iteration 3: loglikelihood = -1980.7961
coef(fit1, matrix = TRUE)
#>             location
#> (Intercept) 1.165758
#> x2          6.095522

# Both location and scale unknown
ldata <- transform(ldata, y2 = rlogis(nn, loc = 1 + 5*x2, exp(x2)))
fit2 <- vglm(cbind(y1, y2) ~ x2, logistic, data = ldata, trace = TRUE)
#> Iteration 1: loglikelihood = -3885.2006
#> Iteration 2: loglikelihood = -3563.3616
#> Iteration 3: loglikelihood = -3349.9904
#> Iteration 4: loglikelihood = -3273.1111
#> Iteration 5: loglikelihood = -3266.0782
#> Iteration 6: loglikelihood = -3266.024
#> Iteration 7: loglikelihood = -3266.0238
#> Iteration 8: loglikelihood = -3266.0238
coef(fit2, matrix = TRUE)
#>             location1 loglink(scale1) location2 loglink(scale2)
#> (Intercept)  1.174509        1.964806  1.052873       0.5701189
#> x2           6.080957        0.000000  4.336689       0.0000000
vcov(fit2)
#>               (Intercept):1 (Intercept):2 (Intercept):3 (Intercept):4      x2:1
#> (Intercept):1      1.291529   0.000000000    0.00000000   0.000000000 -1.913697
#> (Intercept):2      0.000000   0.001398644    0.00000000   0.000000000  0.000000
#> (Intercept):3      0.000000   0.000000000    0.07937843   0.000000000  0.000000
#> (Intercept):4      0.000000   0.000000000    0.00000000   0.001398644  0.000000
#> x2:1              -1.913697   0.000000000    0.00000000   0.000000000  3.713461
#> x2:2               0.000000   0.000000000   -0.11761736   0.000000000  0.000000
#>                     x2:2
#> (Intercept):1  0.0000000
#> (Intercept):2  0.0000000
#> (Intercept):3 -0.1176174
#> (Intercept):4  0.0000000
#> x2:1           0.0000000
#> x2:2           0.2282323
summary(fit2)
#> 
#> Call:
#> vglm(formula = cbind(y1, y2) ~ x2, family = logistic, data = ldata, 
#>     trace = TRUE)
#> 
#> Coefficients: 
#>               Estimate Std. Error z value Pr(>|z|)    
#> (Intercept):1   1.1745     1.1364   1.033 0.301377    
#> (Intercept):2   1.9648     0.0374  52.537  < 2e-16 ***
#> (Intercept):3   1.0529     0.2817   3.737 0.000186 ***
#> (Intercept):4   0.5701     0.0374  15.244  < 2e-16 ***
#> x2:1            6.0810     1.9270   3.156 0.001602 ** 
#> x2:2            4.3367     0.4777   9.078  < 2e-16 ***
#> ---
#> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#> 
#> Names of linear predictors: location1, loglink(scale1), location2, 
#> loglink(scale2)
#> 
#> Log-likelihood: -3266.024 on 1994 degrees of freedom
#> 
#> Number of Fisher scoring iterations: 8 
#> 
#> No Hauck-Donner effect found in any of the estimates
#>