help
81 Subscribers 
help > Pooled two-sample t-tests
Showing 1-2 of 2 posts
Sep 26, 2019 03:09 PM | eas
Pooled two-sample t-tests
Hello,
Thank you for the amazing toolbox, it has been very helpful.
I had a question about using pooled two-sample t-tests in the toolbox...is this possible? Because from what I can tell, you use the unpooled estimate, and there is no option to estimate the variance via the pooled approach.
The reason why I ask is that I need to compare one patient case to a group of control subjects, and according to Rik Henson (working with SPM), he suggests using the pooled two-sample t-test considering a small sample size -> https://www.fil.ion.ucl.ac.uk/~wpenny/drafts/spm-book/anova.pdf
However, I am trying to decide the best way to approach this using the NBS toolbox. Do you have any suggestions?
Any help would be much appreciated! Thanks!
Thank you for the amazing toolbox, it has been very helpful.
I had a question about using pooled two-sample t-tests in the toolbox...is this possible? Because from what I can tell, you use the unpooled estimate, and there is no option to estimate the variance via the pooled approach.
The reason why I ask is that I need to compare one patient case to a group of control subjects, and according to Rik Henson (working with SPM), he suggests using the pooled two-sample t-test considering a small sample size -> https://www.fil.ion.ucl.ac.uk/~wpenny/drafts/spm-book/anova.pdf
However, I am trying to decide the best way to approach this using the NBS toolbox. Do you have any suggestions?
Any help would be much appreciated! Thanks!
Sep 27, 2019 12:09 AM | Andrew Zalesky
RE: Pooled two-sample t-tests
Hi,
The variance type is fixed and cannot be changed unless you alter the code.
Performing group-level inference with a group comprising only one sample is challenging. I am not convinced that changing variance estimation will necessarily solve these challenges.
You may want to consider quantile regression, with the model trained on the healthy control group.The model could then be applied to the single patient.
Andrew
Originally posted by eas:
The variance type is fixed and cannot be changed unless you alter the code.
Performing group-level inference with a group comprising only one sample is challenging. I am not convinced that changing variance estimation will necessarily solve these challenges.
You may want to consider quantile regression, with the model trained on the healthy control group.The model could then be applied to the single patient.
Andrew
Originally posted by eas:
Hello,
Thank you for the amazing toolbox, it has been very helpful.
I had a question about using pooled two-sample t-tests in the toolbox...is this possible? Because from what I can tell, you use the unpooled estimate, and there is no option to estimate the variance via the pooled approach.
The reason why I ask is that I need to compare one patient case to a group of control subjects, and according to Rik Henson (working with SPM), he suggests using the pooled two-sample t-test considering a small sample size -> https://www.fil.ion.ucl.ac.uk/~wpenny/drafts/spm-book/anova.pdf
However, I am trying to decide the best way to approach this using the NBS toolbox. Do you have any suggestions?
Any help would be much appreciated! Thanks!
Thank you for the amazing toolbox, it has been very helpful.
I had a question about using pooled two-sample t-tests in the toolbox...is this possible? Because from what I can tell, you use the unpooled estimate, and there is no option to estimate the variance via the pooled approach.
The reason why I ask is that I need to compare one patient case to a group of control subjects, and according to Rik Henson (working with SPM), he suggests using the pooled two-sample t-test considering a small sample size -> https://www.fil.ion.ucl.ac.uk/~wpenny/drafts/spm-book/anova.pdf
However, I am trying to decide the best way to approach this using the NBS toolbox. Do you have any suggestions?
Any help would be much appreciated! Thanks!