help > Interpret the test statistics
Showing 1-3 of 3 posts
Display:
Results per page:
Jan 20, 2021  12:01 PM | Maximilian Schulz
Interpret the test statistics
Dear Mr. Zalewsky,

first of all I would like to thank you again for this great tool. I have a possibly naive question, since I am still a novice in statistics.

I have used the NBS toolbox to extract a set of functional connections from a large cohort that change significantly with age. Do the individual test statistics tell me anything about how strong these connections change with age, or just if the relation with age is significant. I ask this because I read that the T test in a regression distinguishes whether the slope of the regression line is different from zero and in my logic this also makes a statement about the strength of the relationship.

Best regards

Max
Jan 20, 2021  11:01 PM | Andrew Zalesky
RE: Interpret the test statistics
Hi Max,

the test statistic can give you an estimate of the effect size. For example, Cohen's d measure of effect size can be derived by dividing the t-stat by the sqrt root of the sample size.

Note that the effect size estimate can often be inflated in these kinds of analyses, due to the winner's curse, particularly for small sample sizes.

Andrew

Originally posted by Maximilian Schulz:
Dear Mr. Zalewsky,

first of all I would like to thank you again for this great tool. I have a possibly naive question, since I am still a novice in statistics.

I have used the NBS toolbox to extract a set of functional connections from a large cohort that change significantly with age. Do the individual test statistics tell me anything about how strong these connections change with age, or just if the relation with age is significant. I ask this because I read that the T test in a regression distinguishes whether the slope of the regression line is different from zero and in my logic this also makes a statement about the strength of the relationship.

Best regards

Max
Jan 28, 2021  11:01 AM | Maximilian Schulz
RE: Interpret the test statistics
Thanks Dr. Zalewsky,

that helps a lot, sample size shouldn't be an issue since I am working with almost a thousand subjects.

Best regards

Maximilian Schulz