[NOLAN FIVE]

Apologies for not using code delimiters.

[NOLAN FIVE]

Dear Carlo,

I omitted population instead of governance as the latter is an important variable for my model, while population is not. The correlation coefficients look much better and are lower.


About the time dummies, are they necessary ? Fei mentioned that I should try using them but I do not really see the point in my model.

One last thing I wanted to double check with you,
When I run

xtscc  ln_tb_incidence lnGDPpercapita lnGDPSquared Governance lngovspend DAHmal_3 DAHTB_3 DAHHIV1 Tanzania

Does this provide results for Tanzania only? I thought it did, but Fei said I should be using two country names ... I am bit confused as to why I should do this.

Thank you very much for everything,

Nolan.

[Fei Wang]

Thanks Fei,

I thought that by adding the name of one country 'e.g. Tanzania' it gave me the results for that country only, which is what I want. When I put Ghana and Tanzania together, then which country are those results for?
I have added year dummies and controlled for all of them except one, what should I be looking out for?

Well, Nolan, if you only want to analyze data from Tanzania, you need to use if condition, like "if Tanzania == 1". Please refer to help if. So far, all your results are based on ALL three countries, which is fine as they'll show the average relations across the countries. If you only look at one country, it's time series analysis which means almost all Carlo's and my suggestions are irrelevant.

"I have added year dummies and controlled for all of them except one" -- No, you didn't. You have 18 years which means 17 year dummy variables should be controlled for -- apparently not the case in your regression.

[Carlo Lazzaro]

Nolan:
1) time dummies are relevant in -xtreg,fe-. That said, they do not do harm in -xtscc-, too.
2) your original post was about -xreg-that implies >1 panel of observations. Do all your 3 groups relate to Tanzania?

[NOLAN FIVE]

Ok understood, thank you!

Yes I am running 3 regressions: one for each disease (where my dependant variable is either malaria, TB or HIV), and am doing so for 3 countries: Ghana, Nigeria and Tanzania. I would like to obtain separate results for each country so that I can compare them.

[NOLAN FIVE]

Dear Fei,

I did not realise I had to use time series. What about if I use

xtscc lnTB_incidence Governance lnGDPpercapita lnGDPSquared lngovspending DAHmal_3 DAHTB_3 DAHHIV1 if Tanzania==1

And do so for each country (if Ghana==1 and Nigeria==1) then can I keep my model as it is and consider it panel data?

I am really confused and sincerely hope that what I have just mentioned works !

[Carlo Lazzaro]

Nolan:
1) if you analyze > wave of data for the same country, you have time-series;
2) if you analyze 1 wave of data for many countries, you have cross-sectional data;
3) if you have 1)+2), you have pane data.

At the end of the day, you have:
1) a cross sectional dimension composed of 3 countries (Tanzania; Ghana and Nigeria);
2) a time series-dimension (18 years);
3) the coexistence of 1) and 2) means panel data (and in this scenario a panel data regression outperform three separated regressions);
Hence, since you have a T>N panel dataset, you should go:

xi: xtscc lnTB_incidence Governance lnGDPpercapita lnGDPSquared lngovspending DAHmal_3 DAHTB_3 DAHHIV1 i.year i.country

Then you can test the joint significance of -i.year- via:

xi: testparm i.year 

The -xi:- prefix is necessary here because the community-contributed module -xtscc- does not support -fvvarlist- notation.

[Fei Wang]

-xtscc- may still work, but I'm not sure what the results represent. As you only have a time series data of 18 years, why not just running the following code as a beginning:

reg Y X time_trends if Tanzania == 1, robust

[Carlo Lazzaro]

Fei is obviously correct.
-xtscc- makes sense here if Nolan includes all the countries in the same regression (instead of runnng 3 separate OLS).

[Fei Wang]

Carlo, to my understanding, Nolan only wants separates coefficients by country and compare them among the three countries, as in #20.

[NOLAN FIVE]

Yes that is correct Fei.

Carlo, when I run the code you gave me In #22 then the results are an average of all three countries right?

[Carlo Lazzaro]

Nolan:
they're the coefficients of all the predictirs included in the right-hand side of your regerssion equation adjusted for each other.
You should also have a coefficient for (n-1) countries an (n-1) years.

[NOLAN FIVE]

Yes that is correct. But I am not sure how to interpret these and whether they are enough for my dissertation. Does this provide enough information on the correlation between TB incidence and DAH_TB (foreign aid) for each country ?

Thank you so much

[Carlo Lazzaro]

Nolan:
1) Ok, so you want to go with three different OLS, but because you have 3 different diseases, not because you have three countries (If I got you right);
2) that's why I think that the code suggested in #22 is the way to go (to be repeated for each disease);
3) plese share the coefficients of your -xtscc- regression (for one disease) so that we can take a look at them. Thanks.

[NOLAN FIVE]

Dear Carlo, yes I think that sounds good. Thank you so much for taking the time to help me with this, I really appreciate it.
Example for Malaria:

xi: xtscc lnmal_incidence Governance lnGDPpercapita lnGDPSquared lngovspending DAHmal_3 DAHTB_3 DAHHIV1 i.Year i.Country

xtscc lnmal_incidence lnGDPpercapita lnGDPSquared Governance lngovspending DAHmal_3 DAHTB_3 DAHHIV1 if Tanzania==1

I am not managing to insert the tables as delimited text so I have to use screenshots, sorry.