Correct statistical analyses

#1
Hello,

I have a project I'm working on and I am unsure about which statistical analysis would be correct. My professor is not helping much and I am terrified of failing.

I have two groups that I want to compare their results after treatment. One receives treatment and another group doesn't. Should I use an unpaired t-test or is there a different one?

If I want to examine the likeliness of the illness to come back depending on factors like health history, would I use a binary logistic regression? I want the model to help predict the likelihood of the illness depending on factors.

Finally, if I want to examine the time the illness came back depending on factors like health history, would I use cox proportional regression model? I want the model to help predict the time of the illness coming back depending on factors.

Thank you to anyone who helps!
 

Karabiner

TS Contributor
#2
I have two groups that I want to compare their results after treatment. One receives treatment and another group doesn't. Should I use an unpaired t-test or is there a different one?
I depends on sample size and the level of measurement. If total n > 30 and the dependent variable is interval scaled,
then the Welch-corrected independent samples t-test seems appropriate.
If I want to examine the likeliness of the illness to come back depending on factors like health history, would I use a binary logistic regression? I want the model to help predict the likelihood of the illness depending on factors.
For the prediction of a binary variable by several other variables, multiple binary logistic regression is a good choice.
Finally, if I want to examine the time the illness came back depending on factors like health history, would I use cox proportional regression model? I want the model to help predict the time of the illness coming back depending on factors.
I suppose there are censored cases (no coming back of the illness during the observation period), so this looks
appropriate.

With kind regards

Karabiner
 
#3
Hello,
Thank you so much for your help! I truly appreciate it!

I have one last question: If the sample size is more than 30 and the dependent variable is a nominal variable, what would be the correct test?

I depends on sample size and the level of measurement. If total n > 30 and the dependent variable is interval scaled,
then the Welch-corrected independent samples t-test seems appropriate.

For the prediction of a binary variable by several other variables, multiple binary logistic regression is a good choice.

I suppose there are censored cases (no coming back of the illness during the observation period), so this looks
appropriate.

With kind regards

Karabiner
Thank you once again for your help,
JJ