Hi,

Cross sectional analysis in your case seems inappropriate. I would suggest to consider panel data models combining both cross sections and time series data. This will enable you to do a better analysis.

If you provide me more details about your variables and I can provide more on this.

Kind regards

Thushara

Thanks Thushara, but the study is on only ONE country, not several.

No Arun, it is just a yearly data on mobile subscription from 2006 to 2015 from that country...

No Sir, the data is from World Banks WDI...and meaningful data points are only available for that country from 2006, which prevents me from doing a TS analysis. So I'm wondering if averaging the datasets for the periods 2006-2015 is allowed econometrically? Just don't want to commit any blunder...

Ok, my research question is to understand the interactions of financial access, financial innovation and inclusive growth. The dependent variable (inclusive development) is proxied by Human Development Index, the explanatory variables are: number of bank branches (for financial access), mobile subscription (for financial innovation) and per capita GDP....but I intend doing a ONE country analysis.

...thanks a lot, Arun! I appreciate the time taken to guide me through.

Hi Ngozi, I do appreciate this challenge in your study. The time-period is actually short and the degree of freedom may be too low to produce meaningful results. It appears to me here like you are doing a TS study, I am not too clear about the CS you talk about.  The only way I feel you can resolve the issue is to increase the number of observations by using quarterly data for the variables. I hope you can get the data. The HDI you are using may not adequately measure inclusive growth. The measure of inclusive growth can be computed as Ifeoluwa Ogunriola  is currently doing. I commend you efforts.

Thanks Prof Edo, I'll modify my research question due to lack of data in this case and do a panel data analysis instead...

Dear Ngozi,

I do not understand, why you want to apply CS on the averages (over time) of the countries, thereby giving up a lot of information in your data set. Obviously, you are looking for a relation between variables, which can be assumed to be similar for all countries (otherwise CS would not be feasible). Let me show with a simple example, how you can use all your data in one estimation.

Let us assume the relation is Y=a*X1+b*X2+c, where one can expect similar coeffients a, b and c for all countries in the data set. You have Yst, X1st and X2st, where s is a country number (S countries) and t is a time index (06-15 or 1-10). You can form vectors like Y=(Y1,1,Y1,2,Y1,3,....Y1,10 Y2,1,Y2,2,...................;YS,1,......YS,10) and X1 and X2 in the same way. You can treat these vectors as if thy were time series and estimate the above equation as usual. If you add a time vector T=(1,2,3,....10;1,2,3,...10,1,.....etc) and dummy vectors for some countries or country groups (for which there is reason to assume different coefficients), you can widen the approach by estimating the equation Y=a*X1+b*X2+d*T+fi*Dummyi+c, which could show you, if the relation changes over time and how realistic the similarity over countries really is. This procedure opens also the possibility to analyse lagged effects by estimating with lagged explaining variables.

Before estimating, it is,of course, important to thouroughly analyse the problem on the base of economic theory to find out how the relation should be and how the data should  be transformed.

Thanks Anton, but the analysis is on a SINGLE country (from 1980 - 2015) and not a panel. Maybe my question was misleading. I was thinking of simply averaging the variables over time due to missing values in my main explanatory variable so that I can just do a CS on that country. But I guess that approach will be wrong from all the suggestions I've received so far....