How to determine change in effect over time?

Hello ,

I would like to determine if the effect of binary variable (like gender) on a continuous variable (like a reaction time in ms) can change overtime (between T1 and T2). I have over 100 subjects.

What statistical test could help be answer this question ?

Thank you very much,



TS Contributor
This is represented by the group*time interaction in a "mixed" ANOVA (an ANOVA with a between-subjects-factor and a repeated-measures-factor).

With kind regards

Thank you Karabiner,

I understand your answer, but what in the results of the ANOVA could mean "the effect of gender on reaction time was lowered between T1 and T2" ?

The results of the ANOVA you suggested would not mean : "the time effect on reaction time was different in male and female" ?

What I would like is to be able to say "The effect of age on RT was X at T1 and it was Y at T2", if y is lower than X then the effect is lowered in T2 compared to T1.

Do you get what I mean ?

Thanks so much again


TS Contributor
The effect of gender is the difference between men and women at t1 and t2, respectively.
Whether this effect (the difference between men and women) changes between t1 and t2,
is represented by the time*gender interaction. At the same time, this interaction means that
the difference between t1 and t2 is different for men than for women.

If the interaction is statistically significant, you can have a look at descriptive statistics or at
the graphs, in order to determine the nature of the interaction (e.g. whether the difference
between men and women becomes smaller with time, or larger, or even reverses).

Accordingly, you can use a continuous predictor such as age, and use the age*time interaction
in order to test whther the association between age and outcome is different at t1 than at t2.

With kind regards

Thanks again, it does make a lot of sense. Now my question would be : How to statistically show that the difference at T1 is different than the difference at T2 ?
I know that I can compare male at t1 versus male at t2 versus female at t1 versus female at t2 (in posthoc of the ANOVA). What I seem to get is that male and female are different both at T1 and at T2, but it seems that the gender gap is reduced.

THe estimated at effects of "gender" are significant at both times, but it seems the values of that effect is way smaller at T2.

Thank you so much for helping


TS Contributor
Now my question would be : How to statistically show that the difference at T1 is different than the difference at T2 ?
Within a mixed ANOVA, the interaction between (e.g.) time and group is routinely tested by most statistical packages. A statistically significant interaction time*group would permit to reject the Null hypothesis that the difference between males and females at t1 is the same as the difference between males and females at t2. Whether the difference became smaller or larger, can be seen from the descriptive statstics and/or from graphs.

With kind regards

Last edited:
I see, it seemed just not verry robust to me to state that the difference "looks" bigger or smaller at T2. But if you telling me that it is acceptable that is perfect !thanks so much