HELP predicting dependent variable at various level of an independent variable

[Yuping Yin]

Hello all, I am running a fix-effect model and I'd like to predict dependent variable at various level of mortality and in particular proportionate levels of the mortality.

the code are as follows:

xtset id year

xtreg lgdp lgphysicapital TFP DalysIDA edu_aug_labor2554 edu_aug_labor5564 institutionalquality wfpop,fe cluster(id)

Among them, DALY is a health variable that refers to mortality. I tried margins command to construct a scenario of GDP change by using GDP with a constant mortality rate minus GDP when the mortality rate is reduced to 95%.


margins, at(DalysIDA = gen(DalysIDA))

scalar mean_original = r(b)[1,1]

margins, at(DalysIDA = gen(DalysIDA*0.95))

scalar mean_reduced = r(b)[1,1]

scalar mean_difference = mean_reduced - mean_original

display "The average predicted value of GDP in the original Daly's IDA: " mean_original
display "The predicted average GDP value after reducing Daly's IDA by 5%: " mean_reduced
display "Change in the average predicted value: " mean_difference


However, there are some issues, such as only being able to predict a mean value for the entire model, rather than predicting a value for all observations. Secondly, the predicted value, i.e., GDP, is very small, only $1

Therefore, any suggestions on how to improve the prediction method would be greatly appreciated.

[Carlo Lazzaro]

Yuping:
welcome to this forum.
See if -predict,xb- is what you're looking for.

[Yuping Yin]

Yuping:
welcome to this forum.
See if -predict,xb- is what you're looking for.

Hi Carlo,

Thanks for your welcome and positive response. I tried predict before as follows:

predict lgdp_original, xb

* Create a copy of DalysIDA with a 5% reduction.

clonevar DalysIDA_orig = DalysIDA // keep original value
replace DalysIDA = DalysIDA * 0.95 // reduce 5%

replace DalysIDA = DalysIDA *0.6

* create new scenario
predict lgdp_new, xb

* calculate the change
gen lgdp_change = lgdp_new - lgdp_original

* Restore the original value of Daly's IDA
replace DalysIDA = DalysIDA_orig
drop DalysIDA_orig

*
drop lgdp_original lgdp_new

*results
summarize lgdp_change

gen original_GDP=exp(lgdp_change)

unfortunately, the results still quite small nearly 1$ for each observation. I have no idea on where the problem lies. Do you have any further thoughts on my code?

[Carlo Lazzaro]

Yuping:
as per FAQ, posting an example/excerpt of your data may help. Thanks.