cd "F:\stata与计量经济学\data"
use food, clear
* estimate regression
reg food_exp income
* compute t-critical value
scalar tc975 = invttail(38,.025)
di "t critical value 97.5 percentile = " tc975
* calculating 95% interval estimate
scalar ub2 = _b[income]+ tc975*_se[income]
scalar lb2 = _b[income]- tc975*_se[income]
di "beta 2 95% interval estimate is " lb2 " , " ub2
* examples of computing t-critical values
di "t(30) 95th percentile = " invttail(30,0.05)
di "t(20) 95th percentile = " invttail(20,0.05)
di "t(20) 5th percentile = " invttail(20,0.95)
di "t(30) 97.5th percentile = " invttail(30,0.025)
di "t(30) 2.5th percentile = " invttail(30,0.975)* right-tail test ho:beta2 =0
scalar tstat0 = _b[income]/_se[income]
di "t statistic for Ho: beta2=0 = " tstat0
di "t(38) 95th percentile = " invttail(38,0.05)* using lincom
lincom income
* right-tail test ho:beta2 =5.5
scalar tstat1 =(_b[income]-5.5)/_se[income]
di "t-statistic for Ho: beta2 = 5.5 is " tstat1
di "t(38) 99th percentile = " invttail(38,0.01)* using lincom for calculation
lincom income-5.5* left-tail test ho:beta2 =15
scalar tstat2 =(_b[income]-15)/_se[income]
di "t-statistic for Ho: beta2 = 15 is " tstat2
di "t(38) 5th percentile = " invttail(38,0.95)
lincom income-15* two-tail test ho:beta2 =7.5
scalar tstat3 =(_b[income]-7.5)/_se[income]
di "t-statistic for Ho: beta2 = 7.5 is " tstat3
di "t(38) 97.5th percentile = " invttail(38,0.025)
di "t(38) 2.5th percentile = " invttail(38,0.975)
lincom income-7.5* two-tail test ho:beta1 =0
lincom _cons
* p-value for right-tail test
scalar tstat1 =(_b[income]-5.5)/_se[income]
di "p-value for right-tail test ho:beta2 = 5.5 is " ttail(38,tstat1)* p-value for left-tail test
scalar tstat2 =(_b[income]-15)/_se[income]
di "p-value for left-tail test ho:beta2 = 15 is "1-ttail(38,tstat2)* p-value for a two-tail test
scalar tstat3 =(_b[income]-7.5)/_se[income]
scalar phalf = ttail(38,abs(tstat3))
scalar p3 =2*phalf
di "p-value for two-tail test ho:beta2 = 7.5 is " p3
di "p-value for ho:beta2 = 7.5 is "2*ttail(38,abs(tstat3))* linear combinations of parameters
* estimating a linear combination
estat vce
lincom _cons + income*20* testing a linear combination
lincom _cons + income*20-250
log close
* Appendix 3A Graphing rejection regions
clear
* specify critcal values asglobalsglobal t025=invttail(38,0.975)global t975=invttail(38,0.025)* draw the shaded areas, then draw the overall curve
twoway (function y=tden(38,x),range(-5 $t025)///
color(ltblue) recast(area))///(function y=tden(38,x),range($t975 5)///
color(ltblue) recast(area))///(function y=tden(38,x),range(-55)),///
legend(off) plotregion(margin(zero))///
ytitle("f(t)") xtitle("t")///
text(0-2.024"-2.024", place(s))///
text(02.024"2.024", place(s))///
title("Two-tail rejection region""t(38), alpha=0.05")* one-tail rejection region
twoway (function y=tden(38,x),range(1.6865)///
color(ltblue) recast(area))///(function y=tden(38,x),range(-55)),///
legend(off) plotregion(margin(zero))///
ytitle("f(t)") xtitle("t")///
text(01.686"1.686", place(s))///
title("Right-tail rejection region""t(38), alpha=0.05")* Appendix 3C*set up
clear all*open log
log using app3c, replace text
* define global variables
global numobs 40global beta1 100global beta2 10global sigma 50*set random number seed
set seed 1234567* generate sample of data
set obs $numobs
gen x =10
replace x =20if _n > $numobs/2
gen y = $beta1 + $beta2*x + rnormal(0,$sigma)* regression
quietly regress y x
* test h0: beta2 =10
scalar tstat =(_b[x]-$beta2)/_se[x]
di "ttest of ho b2 = 10 " tstat
* program to generate data and to examine
* performance of interval estimator and* hypothesis test
program chap03sim, rclass
version 11.1
drop _all
set obs $numobs
gen x =10
replace x =20if _n > $numobs/2
gen ey = $beta1 + $beta2*x
gen e = rnormal(0, $sigma)
gen y = ey + e
regress y x
scalar tc975 = invttail($numobs-2,0.025)* calculating 95% interval estimate
return scalar b2 = _b[x]return scalar se2 = _se[x]return scalar ub = _b[x]+ tc975*_se[x]return scalar lb = _b[x]- tc975*_se[x]* calculating t-statistic
return scalar tstat =(_b[x]- $beta2)/_se[x]
end
* display 95% interval for test size with different number
* of monte carlo samples
di "lower bound with 10000 replications is "0.05-1.96*sqrt(0.05*0.95/10000)
di "upper bound with 10000 replications is "0.05+1.96*sqrt(0.05*0.95/10000)
di "lower bound with 1000 replications is "0.05-1.96*sqrt(0.05*0.95/1000)
di "upper bound with 1000 replications is "0.05+1.96*sqrt(0.05*0.95/1000)* simulate command
simulate b2r = r(b2) se2r = r(se2) ubr = r(ub) lbr=r(lb)///
tstatr=r(tstat), reps(10000) nodots nolegend ///
seed(1234567): chap03sim
* display experiment parameters
di " Simulation parameters"
di " beta1 = " $beta1
di " beta2 = " $beta2
di " N = " $numobs
di " sigma^2 = " $sigma^2* count intervals covering true beta2 =10
gen cover =(lbr < $beta2)&($beta2 < ubr)* count rejections of true h0: beta2 =10
gen reject =(tstatr > invttail($numobs-2,0.05))* examine some values
list b2r se2r tstatr reject lbr ubr cover in101/120, table
* summarize coverage and rejection
summarize cover reject
log close
6528
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
