0
$\begingroup$

I am new to statistics. I want to put the evaluation plan of my research. The research will be in the form of a RCT with 2 groups (Intervention and control).

I will be comparing the change of the mean of oral health knowledge before, immediately after, and 6 weeks after the intervention in each group as well as between both groups? Which statistical test will be appropriate?

$\endgroup$
0
$\begingroup$

Assuming normality, to answer the question of oral health change after intervention, you would use a paired t-test separately on the two groups. What you anticipate is to reject the null in the control group and fail to reject null in the control group.

When comparing between groups, then you need to clarify what your null will be. Change of oral health over 6 months between the two groups can be tested using a two sample t-test.

$\endgroup$
  • $\begingroup$ Thank you for your help. Yes, the data follows normal distribution. However, I also want to check the the durability of my intervention. Meaning, I that my design will be 2 (groups: inyervenrion/control)×3(time: before/immediately after/6 weeks after). Should I do multiple t-tests? $\endgroup$ – Salma Elwazeer Dec 21 '18 at 16:37
  • $\begingroup$ Your input is very much appreciated. $\endgroup$ – Salma Elwazeer Dec 21 '18 at 16:43
  • $\begingroup$ Before the intervention, ideally carrying out a 2 sample t-test should not reject the null. That would indicate your two groups are different. For all other scenarios, you can repeat 2 sample t-test to check if means are significantly different between intervention and control groups. $\endgroup$ – Arun Jose Dec 27 '18 at 10:17
0
$\begingroup$

Alternatively, if you do not want to deal with pairs of tests, this can be done with a hierarchical / multilevel model (linear mixed model). The random effect will be the individuals in the group.

$\endgroup$
  • $\begingroup$ I don't know about this. Could you please refer me to a paper/document/video to better understand it? $\endgroup$ – Salma Elwazeer Dec 21 '18 at 16:45
  • $\begingroup$ There are plenty of sources on this site for mixed models. BTW a mixed model is just a linear model with added random effects, what are random effects? Have a look at question for ex. stats.stackexchange.com/questions/4700/… $\endgroup$ – user2974951 Dec 21 '18 at 20:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.