topic Re: Student's T in proc univariate in Statistical Procedures

Student's T in proc univariate

juanvg1972 — Wed, 20 Jul 2016 21:03:05 GMT

Hi,

I would like to understand the meaning of T Student result in a proc univariate.

I am executiing a proc univariate to know the distrubution of a var:

proc univariate data=tabsas.consumo normal plot;
var hc_consumo;
histogram hc_consumo/normal (color=red W=5);
run;

In the resullts get:

Tests : Mu0=0
T Student 33,00654 p valor < 0,001

I don't have a statistics background ; i have a basic level, Can anybody helep me??

I have read about T Student, but I don't understand the relation with the result I get...

Any advice will be greatly appreciated

Thanks

Re: Student's T in proc univariate

Reeza — Wed, 20 Jul 2016 21:24:21 GMT

It's a hypothesis test, where the hypothesis is that Mu0 is 0.

Read ad up on basic hypothesis testing here:

https://onlinecourses.science.psu.edu/stat200/node/54

And find annotated output for proc univariate here:

http://www.ats.ucla.edu/stat/sas/output/univ.htm

Additionally, the statistics in SAS e course is free and on demand.

http://support.sas.com/training/tutorial/

Re: Student's T in proc univariate

juanvg1972 — Wed, 20 Jul 2016 21:42:03 GMT

I understand that you make this test, to know is the mean value of the var is represetative, if it represents well the values of the var, in't it??

Re: Student's T in proc univariate

Reeza — Wed, 20 Jul 2016 21:45:11 GMT

No, it's a test to see if the mean value is different than 0.

Re: Student's T in proc univariate

juanvg1972 — Wed, 20 Jul 2016 21:50:32 GMT

Ok, there must be something I don't understand. If I get a mean value of 35,6 (fo9r example , Why do I do a test to know if it is different from 0?

I don't undrestand the objective of this test..., there must be something I don't get..

Re: Student's T in proc univariate

Reeza — Wed, 20 Jul 2016 22:11:46 GMT

Because the means is a single measure. You DON'T KNOW that it is actually different than 0.

Here's an example, extreme, but illustrates the issue of an outlier.

data sample;
	do i=1 to 10;
		x=rand('normal', 0, 2);
		output;
	end;
	x=350;
	output;
run;

proc univariate data=sample;
	var x;
run;

Output:

Tests for Location: Mu0=0
Test	Statistic		p Value
Student's t	t	1.002908	Pr > \|t\|	0.3396
Sign	M	0.5	Pr >= \|M\|	1.0000
Signed Rank	S	7	Pr >= \|S\|	0.5771

In this case the test tells you the mean is not significantly different than 0, although the mean is ~32.

Re: Student's T in proc univariate

Reeza — Wed, 20 Jul 2016 22:12:56 GMT

Anscombe's Quartet is a good example of relying on statistics.

https://en.wikipedia.org/wiki/Anscombe%27s_quartet

Re: Student's T in proc univariate

PriyaL — Thu, 15 Sep 2016 10:04:14 GMT

Hi Reeza,

Would you know if it is possible to adjust the default mean (mu0) value from 0 to my own hypothetical mean as I want to test if my sample lie close to a hypothetical mean using the t-test.

Thanks

Priya.