topic Re: When I use proc univariate for hypothesis testing the t value and p values do not make sense in Statistical Procedures

When I use proc univariate for hypothesis testing the t value and p values do not make sense

jtflynn33 — Mon, 29 Jan 2018 05:12:09 GMT

The data comes from the data table: marchflights.sas7bdat (airplane flights and the corresponding weights of "freight" - the variable)

Freight (variable)

Mean=333.95, Standard deviation=98.11. Number of observations=635

Below is the code:

/**Hypothesis Testing**/
proc univariate data=certprep.marchflights mu0 = 333.95;
var freight;
run;

I believe that the t statistic should be much higher than 0.000283,

and the p value should be very low, looking like 0.0040 or something like that, but it is 0.9998

Results are below:

Tests for Location: Mu0=333.95

0.000283	0.9998
t statistic	p value

Can the results of the t statistic and the p value actually be correct? I don't think so.

Is there anything I'm missing here? Is there something wrong with the code?

Thank you.

jtflynn33

Re: When I use proc univariate for hypothesis testing the t value and p values do not make sense

Ksharp — Mon, 29 Jan 2018 12:41:42 GMT

It looks right. If you want p value be small (significant) , then t statistic must be big enough.

t statistic measured the deviation of H0 statistical (when H0 is assuming right) .

Re: When I use proc univariate for hypothesis testing the t value and p values do not make sense

PaigeMiller — Mon, 29 Jan 2018 13:46:27 GMT

t = (xbar - mu) / (s / sqrt(n))

xbar and mu are both 333.95 (well, xbar probably has a few more decimal places), so t close to zero is correct.

When you have a t close to zero, you have a p value close to 1

Re: When I use proc univariate for hypothesis testing the t value and p values do not make sense

Reeza — Mon, 29 Jan 2018 16:43:21 GMT

I think you're misunderstanding hypothesis testing.

The null hypothesis is that the mean is equal to 333.95. If p<Alpha then you would reject the null hypothesis.

Since the P-value is 1, this means you fail to reject the null hypothesis and the mean is not significantly different than 333.95.

Usually you're looking to reject a null hypothesis, in this case you're looking to confirm your null hypothesis.

@jtflynn33 wrote:

The data comes from the data table: marchflights.sas7bdat   (airplane flights and the corresponding weights of "freight" - the variable)

Freight (variable)

Mean=333.95, Standard deviation=98.11. Number of observations=635

Below is the code:

/**Hypothesis Testing**/
proc univariate data=certprep.marchflights mu0 = 333.95;
     var freight;
run;

I believe that the t statistic should be much higher than 0.000283,

and the p value should be very low, looking like 0.0040 or something like that, but it is 0.9998

Results are below:

Tests for Location: Mu0=333.95

0.000283 0.9998

t statistic   p value

Can the results of the t statistic and the p value actually be correct? I don't think so.

Is there anything I'm missing here? Is there something wrong with the code?

Thank you.

jtflynn33

Re: When I use proc univariate for hypothesis testing the t value and p values do not make sense

jtflynn33 — Sat, 03 Feb 2018 16:02:03 GMT

Thanks, your answer really helped me.

Re: When I use proc univariate for hypothesis testing the t value and p values do not make sense

jtflynn33 — Sat, 03 Feb 2018 16:03:44 GMT

Thanks for your help!!