#4-#6 on Homework 3

Here are the codes I used for #4-#6 on Homework 3 (you have to do #3 to do #4 so I included it). I think the graph in #6 could be better, but it is a start.

3.
. generate logdol = log(dollars)

. generate logincid = log(incid)
(4 missing values generated)

. generate logprev = log(preval)
(4 missing values generated)

. generate loghospdays = log(hospdays)

. generate logmort = log(mort)

. generate logyrslost = log(yrslost)

. generate logdisbil = log(disabil)

. sw regress logdol logincid logprev loghospdays logmort logyrslost logdisbil, forward pe(.1) pr(.2)


4. . sw regress logdol logincid logprev loghospdays logmort logyrs
> lost logdisbil, pe(.1) pr(.2)

5. regress logdol loghospdays


6. regress logdol loghospdays logmort logyrslost logdisbil


. predict yhat, xb

.predict h, leverage

. predict std_yhat
(option xb assumed; fitted values)

. predict std_f, stdf

. generate ciu_f = yhat + invttail(25-5-1, .025)*std_f

. generate cil_f = yhat - invttail(25-5-1, .025)*std_f

. generate cil_sbpf = exp(cil_f)

. generate ciu_sbpf = exp(ciu_f)

. display invttail(19, 0.025)
2.0930241

. predict t, rstudent

lowess t yhat, bwidth(0.8) mcolor(gs10) clwidth(thick) yline(-1.96 0 1.96) x
> label(4.7(.1)5.1)

Chapter 2, #8

* Question 8

use "(your data source)\2.ex.vonHippelLindau.dta"

codebook disease

generate logp_ne = ln( p_ne)

generate logtumorvol = ln( tumorvol)

* VonHippel-Lindau regression

regress logp_ne logtumorvol if disease == 0

regress logp_ne logtumorvol if disease == 1

* Slope estimate for vonHippel-Lindau = .242; the 95%CI = (.056, .428)

* Slope estimate for Multiple Endocrine Neoplasia = .242, 95%CI = (.196, .671).

generate s2 = (.374688289*26 + .267608258*7)/(28+9-4)

generate var_dif = s2*(.0903891^2/.374688289+.1003964^2/.267608258)

generate t = (.2418597-.4337545)/sqrt(var_dif)

generate ci95_lb = (.2418597-.4337545) - invttail(33, .025)*sqrt(var_dif)

generate ci95_ub = (.2418597-.4337545) + invttail(33, .025)*sqrt(var_dif)

list s2 var_dif t ci95_lb ci95_ub in 1/1

display 2*ttail(33, abs(t))

* The null is confirmed. These two slopes are statistically equal.

* The 95% CI = (-.486, .102)

Code for Ch.2 #7

use "(your data source here)\2.ex.vonHippelLindau.dta"

generate logp_ne = ln( p_ne)

generate logtumorvol = ln( tumorvol)

regress logp_ne logtumorvol

predict lnresid, rstudent

predict logyhat, xb

predict logstdp, stdp

generate logci_u = logyhat+invttail(_N-2, .025)* logstdp

generate logci_l = logyhat-invttail(_N-2, .025)* logstdp

regress logp_ne logtumorvol

display invttail(_N-3, .025)

scatter lnresid logtumorvol, yline(-2.0322445 0 2.0322445) lowess lnresid logtumorvol

Code for #11, Chap 2

generate logsbp=log(sbp)

generate logbmi=log(bmi)

regress logsbp logbmi if sex==1

predict yhatmen, xb

predict menstd_f, stdf

generate menci_uf = yhatmen + invttail(_N-2,0.025)* menstd_f

generate menci_lf = yhatmen - invttail(_N-2,0.025)* menstd_f

scatter logsbp logbmi if sex==1, msymbol(o) scatter yhatmen logbmi if sex
> ==1, c(l) s(i) scatter menci_uf logbmi, c(l) s(i) scatter menci_lf log
> bmi, c(l) s(i)

regress logsbp logbmi if sex==2

predict yhatwom, xb

predict womstd_f, stdf

display invttail(_N-2,0.025)
1.9604692

generate womci_uf = yhatwom + invttail(_N-2,0.025)* womstd_f

generate womci_lf = yhatwom - invttail(_N-2,0.025)* womstd_f

scatter logsbp logbmi if sex==2, msymbol(o) scatter yhatwom logbmi if sex
> ==2, c(l) s(i) scatter womci_lf logbmi, c(l) s(i) scatter womci_uf logb
> mi, c(l) s(i)

Welcome to the blog!

I created this blog so that we could post our STATA problems. I thought this would be much easier than trying to copy it all down in class. I think this will help us get through our class.