help
help > RE: 2x2 ANOVA setup
Mar 28, 2020 01:03 PM | Alfonso Nieto-Castanon - Boston University
RE: 2x2 ANOVA setup
Hi Amy,
CONN will use T-stats for ANOVAs with two levels for each factor and F-stats for ANOVAs with three or more levels (for details about GLM stats see https://web.conn-toolbox.org/fmri-method...). In reality that is somewhat arbitrary since two-sided T-stats and F- stats are exactly equivalent (and the only reason to use T- instead of F- values in this case is just to also allow directional tests). If you want to transform T-stats to F-stats simply use the equivalence:
T(df) ^2 = F(1,df)
So, for example, if CONN reports in voxel-based analyses |T(15)|>3, you may simply report instead F(1,15)>9; or if in an ROI-to-ROI analysis you are obtaining "T(15) = 2, p = 0.0639 (two-tailed)", you may equivalently report that as "F(1,15) = 4, p =0.0639"
Best
Alfonso
ps. you may check this equivalence with the syntax:
T=randn; % T-stat value
df=randi([1 10]); % degrees of freedom
p1=1-spm_Tcdf(T,df); % one-sided p-value (T-stat)
p1=2*min(p1,1-p1); % two-sided p-value (T-stat)
p2=1-spm_Fcdf(T^2,1,df); % p-value (F-stat)
disp([p1 p2])
Originally posted by Amy Bouchard:
CONN will use T-stats for ANOVAs with two levels for each factor and F-stats for ANOVAs with three or more levels (for details about GLM stats see https://web.conn-toolbox.org/fmri-method...). In reality that is somewhat arbitrary since two-sided T-stats and F- stats are exactly equivalent (and the only reason to use T- instead of F- values in this case is just to also allow directional tests). If you want to transform T-stats to F-stats simply use the equivalence:
T(df) ^2 = F(1,df)
So, for example, if CONN reports in voxel-based analyses |T(15)|>3, you may simply report instead F(1,15)>9; or if in an ROI-to-ROI analysis you are obtaining "T(15) = 2, p = 0.0639 (two-tailed)", you may equivalently report that as "F(1,15) = 4, p =0.0639"
Best
Alfonso
ps. you may check this equivalence with the syntax:
T=randn; % T-stat value
df=randi([1 10]); % degrees of freedom
p1=1-spm_Tcdf(T,df); % one-sided p-value (T-stat)
p1=2*min(p1,1-p1); % two-sided p-value (T-stat)
p2=1-spm_Fcdf(T^2,1,df); % p-value (F-stat)
disp([p1 p2])
Originally posted by Amy Bouchard:
Hello Alfonso,
Could you please tell me how would one obtain F values for the contrasts 1) main effect of group, 2) main effect of condition, 3) interaction between group and condition?
I have the same design (2 x 2 ANOVA), however, in the 2nd level results, it shows a t-statistic instead of an F-statistic.
Thanks,
Amy
Could you please tell me how would one obtain F values for the contrasts 1) main effect of group, 2) main effect of condition, 3) interaction between group and condition?
I have the same design (2 x 2 ANOVA), however, in the 2nd level results, it shows a t-statistic instead of an F-statistic.
Thanks,
Amy
Threaded View
Title | Author | Date |
---|---|---|
Greg Book | Mar 12, 2014 | |
Alfonso Nieto-Castanon | Apr 14, 2014 | |
Victor Pando-Naude | Feb 1, 2023 | |
Amy Bouchard | Mar 27, 2020 | |
Alfonso Nieto-Castanon | Mar 28, 2020 | |
Amy Bouchard | Mar 28, 2020 | |
Bruno Baumann | May 2, 2016 | |
Lars Michels | Oct 2, 2015 | |
Alfonso Nieto-Castanon | Oct 5, 2015 | |
Lars Michels | Oct 6, 2015 | |
Alfonso Nieto-Castanon | Oct 8, 2015 | |
Lars Michels | Oct 8, 2015 | |
Alfonso Nieto-Castanon | Oct 9, 2015 | |
Greg Book | Mar 31, 2014 | |