Programming the statistical procedures from SAS

T-values for averages?

Reply
Contributor
Posts: 45

T-values for averages?

I have t-values for mean returns and actual values of mean returns fo r20 countries, and have calculated averages of the country returns. Is it possible based on the  t-values I have to calculate t-values also for the average returns, or do I have to code it earlier?

Grand Advisor
Posts: 9,463

Re: T-values for averages?

I don't understand what t-value you mean .

Do you get it from proc ttest ?

Contributor
Posts: 45

Re: T-values for averages?

Just the t-statistic I get from testing whether the returns are statistically different from zero. Using:

proc means mean t - approach. So I'm just wondering if I can take an average of the t-statistics as well, or is it statistically impossible?

Grand Advisor
Posts: 9,463

Re: T-values for averages?

I think you should firstly test your data (return) to see whether it conform NORMAL distribution, because t is based on NORMAL distribution, and also should use non-parameter method to test it again to make the result is reliable .

Ksharp

Contributor
Posts: 45

Re: T-values for averages?

I've done normality testing using proc univariate normal, and the Shapiro-Wilk values are on 99% of the cases over 0.90. Most are somewhere close to 0.98. I think this is enough to know that it's close to normal, right? Am I able to get the t-statistic averages now somehow?

Grand Advisor
Posts: 9,463

Re: T-values for averages?

By the way, These return is time serial data ?

Normal Test required these data (i.e. observations) is independent for each other .

So I am not quite sure if it is right to do so. Maybe you need some statistician ,especially for the time serial analysis .

Respected Advisor
Posts: 2,655

Re: T-values for averages?

Ksharp is trying to get you on the right track here.  If this is time series data, getting a whole bunch of t values, say one for each time point, and then averaging them falls into the category of a very bad idea.  First, the values are not independent and have some sort of serial correlation. Second, how could you possibly calculate the correct degrees of freedom to come up with a p value or standardized effect size?  I think you need to look at the ETS suite, especially PROC PANEL, if this describes your dataset.

Try this for a start--look at PROC MIXED for a repeated measures analysis, with country and time as CLASS variables.  Your model should include the interaction between these to get separate estimates by country and time.  I don't know what the repeated nature of the time variable is like, but an AR(1) structure probably fits if your data are equally spaced in time.  It may be heterogeneous by country, so that is something else you will have to investigate.  In any case, the Type III tests are "averaged t-tests" in this context.  Specific t values can be obtained with an LSMEANS statement.

Good luck.

Steve Denham

Ask a Question
Discussion stats
  • 6 replies
  • 210 views
  • 0 likes
  • 3 in conversation