[Gretl-users] Re: How to change the reference category in multinomial logit

On 3 Jul 2019, at 4:37 pm, Riccardo (Jack) Lucchetti <r.lucchetti(a)univpm.it&gt; wrote: On Wed, 3 Jul 2019, Fazeel Jaleel wrote: > Hello, > Can someone please explain , > 1.how to change the reference category in multinomial logic estimation in Gretl > 2. And, how to obtain estimates for each category for an explanatory variable that has 4 categories. At present, the "logit" command doesn't have an option for choosing the reference category when estimating the multinomial model; the reference category is always the lowest value. However, you can use the "recode" function to redefine the dependent variable to make the reference category the one you want. Once estimation is done, the fitted probvabilities for each category can be recovered via the "$mnlprobs" accessor. All this is exemplified in the script below: <hansl> # Replicate the multinomial logit example from Table 15.2 in # Wooldridge's panel data book (2002a). open keane.gdt smpl (year==87) --restrict list X = 0 educ exper expersq black logit status X --multinomial # change reference category status2 = replace(status, {1,2}, {2,1}) xtab status status2 logit status2 X --multinomial # retrieve fitted probabilities matrix P = $mnlprobs print P --range=1:10 </hansl> Hope this helps! ------------------------------------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti -------------------------------------------------------_______________________________________________ Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/