topic Re: T-Test results with Null Values in Statistical Procedures

T-Test results with Null Values

Lost_Gary — Fri, 28 Jun 2019 18:38:22 GMT

I am performing proc ttest and attempting to compare the equal and unequal variances. Some of my unequal variances are missing. I assume that this is because of very small observations within the sample (such as only 1). My question is how do I interpret this within the ttest procedure. Should i refer to the equal variances or consider these to small to calculate?

Thanks

Re: T-Test results with Null Values

Reeza — Fri, 28 Jun 2019 19:13:45 GMT

Too small to calculate

Re: T-Test results with Null Values

PaigeMiller — Fri, 28 Jun 2019 19:17:15 GMT

I have a different opinion. The fact that you have only one data point for some level of the category variable in the t-test does not mean that the variances are too small to calculate. It means that the variances cannot be calculated.

Re: T-Test results with Null Values

Reeza — Fri, 28 Jun 2019 19:20:50 GMT

Variances don't exist for an N=1 🙂

Re: T-Test results with Null Values

PaigeMiller — Fri, 28 Jun 2019 19:22:32 GMT

@Reeza wrote:
Variances don't exist for an N=1 🙂

Population variances do exist, but you can't estimate them with the sample of 1.