testing normality

[ralphbenno]

I have 2 variables   male, female

and mortality rate

I use proc ttest to test the NULL

do I use proc mix to test for normality

and If normality is broken then we need to use the willcoxon rank sum test?

can you help me with the order.>

below is my code

F test;
proc ttest data = Project;
class sex;
where sex in ("f" "m");
var rate;
run;
quit;

*check residue for normality;
proc mixed data = Project;
class sex;
where sex in ("f" "m");
model rate = sex/residual;
run;
quit;

*wilcoxom rank sum;
proc npar1way data = Project  wilcoxon;
class sex;
where sex in ("f" "m");
var rate;
run;
quit;

I do not need to do a tukey test for pair. right?

* Tukey;
proc glm data=project;
class sex;
where sex in ("f" "m");
model rate = sex / solution;
means sex / tukey;
run;
quit;

[Reeza]

T-Test is fairly resilient into devotions from normality, usually it's more important to ensure that the distribution is the same between the variables. This is more true as N gets larger.

[ralphbenno]

then if I use proc ttest, I do not need to do proc mixed test?

[Reeza]

---

@ralphbenno wrote:

then if I use proc ttest, I do not need to do proc mixed test?

---

I've never used PROC MIXED so no idea.

[Ksharp]

If your class variable has only two levels , TTest is good enough,

If you worry about whether rate is normal distribution or not, then wilcoxom rank sum test is better choice.