Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Forecasting

From   "Victor Zammit" <>
To   <>
Subject   st: Forecasting
Date   Mon, 14 Nov 2011 15:09:15 +0100

* Dear Statalist,

* Using auto.dta,I want to determine the rank of the relative predictive faculty of the following nine

* independent variables, ( price, headroom, trunk, weight, length, turn, displacement ,gear_ratio, foreign).

* All of the nine models are atatistically significant at 95% confidence,because in all

* cases,the t-value is above 2.00 standard deviations. Can anyone please advice me whether I am

* on the right path? Also,is it coincidental,that the R-squared ranking matches perfectly that of the

* and the absolute t-values ranking.What is the intuition ,of the very high correlation,.9954,between

* the R-squares and the respective absolute values of the t-values ?

* Victor M Zammit

version 11

capture program drop kusi

program define kusi

use auto,clear

sort `1'

reg mpg `1'

predict a

line a mpg `1',saving(a`2',replace)

gen str12 variable = "`1'"

gen r2 = e(r2)

gen t_value = _b[`1']/_se[`1']

gen abs_val = t_value

replace abs_val = t_value*-1 if t_value<0

keep t_value r2 variable abs_val

keep in 1

save a`2',replace


kusi price 1

kusi headroom 2

kusi trunk 3

kusi weight 4

kusi length 5

kusi turn 6

kusi displacement 7

kusi gear_ratio 8

kusi foreign 9

graph combine a1.gph a2.gph a3.gph a4.gph a5.gph a6.gph a7.gph a8.gph a9.gph

use a1,clear

forvalues i = 2/9 {

append using a`i'.dta


gen s = 1-r2

sort s

drop s


correlate abs_val r2

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index