ttest2 with a group describing independent values (statistics)

Hi,

I use Matlab ttest2 function to assess if observations from vectors x and y come from populations with different means. The function works fine if vectors are passed as arguments:

   [h, p] = ttest2(x, y)

My data includes biomarker (as blood pressure) in x and y during day and night, respectively. Therefore, we assess paired groups into a matrix [x y], where:

x = [120, 110.5, 123, 140, 142, 151, 121, 119, 110, 115, 130]'  % there are more observations
y = [118, 110, 120, 135, 134, 140, 120, 116, 110, 113 ,127]'    % there are more observations
m = [x y];

The point is that we also want to compare unpaired or independent groups (males vs. females), at day and night. For this, a third column is included defining the group (male=0, female=1) in matrix [x y g], where

g = [0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 1]'   % there are more observations
m = [x y g];

Function anova1() has the functionality of including a group as second argument for each observation in x or y.

I wonder if it is possible to use ttest2 function in anova1() fashion, something like

[h, p] = ttest2(x, g)   % where x contains the observation and g contains categorical values

To compare the biomarker, e.g. during night, between male and female, without rearrenge vectors

Thanks

Answer 1

[Jeff Miller]

No, it is not possible to use ttest2 as you suggest.

Your design has 2 factors: gender and time (day/night) with repeated measures on the time factor, and you can conveniently analyze it with 'fitlm' and 'ranova', something like this (note the categorical conversions added on edit):

x = [120, 110.5, 123, 140, 142, 151, 121, 119, 110, 115, 130]';  % there are more observations
y = [118, 110, 120, 135, 134, 140, 120, 116, 110, 113 ,127]';    % there are more observations
g = [0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 1]';   % there are more observations
m = [x y g];
t = array2table(m,'VariableNames',{'x','y','gender'});
t.gender = categorical(t.gender);
timeDesign = table([1 2]','VariableNames',{'time'});
timeDesign.time = categorical(timeDesign.time);
rm = fitrm(t,'x-y~gender','WithinDesign',timeDesign);
ranova(rm,'WithinModel','time') 

[Pedro Gomis]

Thanks for your answer!

I will try fitrm and repeated measures anova.