[atta asif]

thanks, Farnando
is it possible for you to guide me so that i may work on it myself, as my research work is based on this methodology?

[FernandoRios]

unfortunately not.
Im not familiar with Bayes Econometrics. And my commands were not written to work with bayes.

[atta asif]

Hi, Fernando hope you are doing great.
i need your help about the robustness checks. i have used the RIF command you have suggested/guided me but a reviewer comments that you need to run some robsutness checks, so please guide me in this regard. how should i respond to her/him?

thanks
Atta

[FernandoRios]

Hi atta
your question is very vague
I think it would depend on what you did exactly, and what are other comments from the reviewer

[atta asif]

i mean i have ran the RIF regression and Oaxaca RIF commands to find estimates of different socioeconomic factors such as regions, wealth status, parental education etc. now want to check robustness of this model. please can you guide me on how can do this?

[atta asif]

i mean i have ran the RIF regression and Oaxaca RIF commands (righdreg and oaxaca_rif) to find estimates of different socioeconomic factors such as regions, wealth status, parental education etc. now want to check robustness of this model. please can you guide me on how can do this?

[Duong Le]

Hi Atta
I see. If I understand the process correctly, it would have to be something as follows:

webuse nhanes2f, clear
oprobit health female black age c.age#c.age
predict xb_pr,xb
sum xb_pr
gen rchealth=(xb_pr-r(min))/(r(max)-r(min))
rifhdreg rchealth female black age c.age#c.age, rif(eindex(rchealth) lb(0) ub(1) )

the only difference would be that within the option "rif" you will need to add the income or wealth variable instead of the RCHEALTH variable).

I also still read carefully the paper they cite:
Van Doorslaer E, Jones AM. Inequalities in self-reported health: validation of a new approach to measurement. J Health Econ. 2003;22:61–87.
As it may explain alternatives to how to implement the initial transformation.

Best

Dear FernandoRios,

It's always great to see your posts on decomposition. I have learnt lots from those posts. I have a related question to this thread. I am conducting a comparative study on socioeconomic-related health inequality in psychological well-being in two countries. However, questions related to psychological well-being in the two countries are not the same, say, country A uses a set of 5 questions to measure psychological well-being, while country B uses the same 5 questions as those of country A, but also add 2 new more (in total 7 questions for country B). Each of 5/7 questions has 3-scale responses (no, a bit, and more) so I recode those responses as 0=no, 1=a bit; and 2=more. As a result, the psychological well-being score in country A ranges from 0-10 and that is 0-14 in country B.

I have three questions that I want to have advice from you.
1) do I need to run your code (please see my quote) to get a recenter psychological index? I am thinking of another way but not sure if it is appropriate. My approach is as follows

* psy_A: psychological well-being score in country A
* psy_B: psychological well-being score in country B
// For country A
sum psy_A
sca max_psyA = r(max)
sca min_psyA = r(min)
gen rcA = (psy_A - min_psyA)/(max_psyA - min_psyA)

// For country B
sum psy_B
sca max_psyB = r(max)
sca min_psyB = r(min)
gen rcB = (psy_B - min_psyB)/(max_psyB - min_psyB)

By doing so I also can normalize psy_A and psy_B to a new re-scale interval (0, 1). Do you think that now the psychological well-being scores (interval one) between the two countries are comparable? because the values of both of them range from 0-1 instead of from 0-10 and 0-14.

2) after rescaling, am I good to perform RIF decomposition? given a measure of SES/income is available and well-defined.
3) In case, you may think that running a regression and obtain predicted values is needed, can I treat psy_A and psy_B as continuous variable and just run OLS instead of oprobit?

Thank you.

[FernandoRios]

Hi Jason,
First of all, declaimer, Some of your questions are a bit outside my field.
so:
1. From a mathematical point of view, it seems reasonable to compare RcA and rcB because they are both measured in a scale from 0 to 1.
Perhaps i would also try to create a second measure that uses the same questions for both countries. That seems to be more consistent, because the same factors are used across countries before you do the decomposition. Otherwise, an addition reason why their "welfare" measure differ will be because they ask different number of questions.

2. Using RIF decomposition on the index seems reasonable too. Although I haven't read anything on that particular subject (most of the decompositions i read about are related to wages and wealth). The question will be how do you interpret the rescaled coefficient. What will 0 be or what will 1 be? how is the index usually interpreted in the literature?
3. I dont think you need to do that. I remember that the reason to run an oprobit before was to obtain a predicted index that could be used for decomposition. You will have to consult the paper cited earlier in this thread. I dont think you need to do that step. Unless other people in your field have done the same.

HTH

Fernando

[Duong Le]

Hi Jason,
First of all, declaimer, Some of your questions are a bit outside my field.
so:
1. From a mathematical point of view, it seems reasonable to compare RcA and rcB because they are both measured in a scale from 0 to 1.
Perhaps i would also try to create a second measure that uses the same questions for both countries. That seems to be more consistent, because the same factors are used across countries before you do the decomposition. Otherwise, an addition reason why their "welfare" measure differ will be because they ask different number of questions.

2. Using RIF decomposition on the index seems reasonable too. Although I haven't read anything on that particular subject (most of the decompositions i read about are related to wages and wealth). The question will be how do you interpret the rescaled coefficient. What will 0 be or what will 1 be? how is the index usually interpreted in the literature?
3. I dont think you need to do that. I remember that the reason to run an oprobit before was to obtain a predicted index that could be used for decomposition. You will have to consult the paper cited earlier in this thread. I dont think you need to do that step. Unless other people in your field have done the same.

HTH

Fernando

Dear Fernando,

Thank you so much for your comments and suggestions. They are very helpful to my research.

Best,

Jason