Replicate the Worldwide Governance Indicators methodology to construct my own index

[Canh Dang]

Hi!

I am a Master student who want to construct my own accountability index, using a dataset from an African country in non-profit sectors. I have around 200 organisations, and around 25 questions regarding 4 aspects of accountability (so around 6 questions per each). I have several questions and I greatly appreciate any help.

1) Can anyone please refer some economics papers that the authors also construct their own indices/measurements? Is it acceptable for a master student to do so in his paper?

2) I follow the Worldwide Governance Indicators methodology, where I consider each aspect of accountability as a data sources (as in the WGI). As in the link, each aspect will be k, and each organisation will be j. I have already the observed index y_j and I will use the unobserved component model to calculate the g_j (please refer to the link to see the notation).

I am a beginner in stata programming so I want to ask for help to programme the maximum likelihood estimator in the Appendix (I upload the screenshot of the equation).

My question is twofold. Can I follow the methodology and can you please give me some warnings. As it is my first ever paper, I am really excited about the idea but when I start working on it, I feel the pressure.

Second, can anyone please provide me some guidance on how to programme the ML estimators.

Thank you very much.

[Canh Dang]

I try to edit the post but I can't. I am sorry for the double posts.

I want to programme in Stata the problem below:

[Phil Bromiley]

I'm afraid you are asking for more than this group normally does. We do not do literature searches for students (or other scholars). If people happen to know of a paper, they might mention it, but finding papers in economics that create indices is your job. There are large literatures on creating indices in many disciplines including psychology, sociology, and economics. Get on JSTOR or a similar site, and start searching.

We also don't normally help at this unformed stage with estimation. If you had an almost working program, we'd probably help you. If someone knew of a pre-packaged program that did the estimate you want, s/he might mention it. But throwing out a general statement of a likelihood function and asking us what to do is way too much to ask.

Given what appears to be your level of skill, in my opinion, you would probably be better advised to work with pre-programmed routines rather than trying to program a maximum likelihood estimator from scratch.