# Basic Stats - T-test, ANOVA, Chi sq?

#### teddybear

##### New Member
Hi all,

I am obviously no stats major but am having trouble with the analysis in my research. I am basically looking at the mean BMI (Normal, Overweight, Obese) in three different groups of individuals (Heart attack patients, Neurological patients, and Other).

Which test should I do to
1) see the differences in mean BMI across the groups of individuals
2) See the differences between mean BMI within just the heart attack patients

Thank you!

edit: if anyone knows how to do this on SAS as well, bonus points

#### obh

##### Well-Known Member
Hi all,

I am obviously no stats major but am having trouble with the analysis in my research. I am basically looking at the mean BMI (Normal, Overweight, Obese) in three different groups of individuals (Heart attack patients, Neurological patients, and Other).

Which test should I do to
1) see the differences in mean BMI across the groups of individuals
2) See the differences between mean BMI within just the heart attack patients

Thank you!

edit: if anyone knows how to do this on SAS as well, bonus points
1. I assume you want to check if the difference between the average BMI across the groups is significance?
2. What do you want to test here?

#### ooostats

##### Member
Also, you mention that you are looking at "the mean BMI (normal, overweight, obese)". BMI is a continuous variable, whereas the classifications of BMI into normal/overweight/obese means you treat it as a categorical variable. In either case, you should be clear as to what level of measurement your dependent variable is on as this directly informs the test that you do.

#### teddybear

##### New Member
1. I assume you want to check if the difference between the average BMI across the groups is significance?
2. What do you want to test here?

Hi yes exactly. For 2 I would want to see if there is a difference between normal weight, overweight, and obese patients in just heart attack patients.

#### teddybear

##### New Member
Also, you mention that you are looking at "the mean BMI (normal, overweight, obese)". BMI is a continuous variable, whereas the classifications of BMI into normal/overweight/obese means you treat it as a categorical variable. In either case, you should be clear as to what level of measurement your dependent variable is on as this directly informs the test that you do.
Sorry, the mean BMI of each categorical group (Underweight/normal, Overweight, Obese)

#### Karabiner

##### TS Contributor
Do you mean, for example, that you want to know whether
those neurological patients who have a BMI between 25.0 to <30
(overweight) have a higher mean BMI than those heart attack patients
who have a BMI between 25.0 to <30? Which underlying research
question will be answered by this?

And, by the way, the crucial information is missing (sample size).

As to question 2) I am not sure what I might be missing:
2) See the differences between mean BMI within just the heart attack patients
It would make no sense to test whether normal/overweight/obese
differ with respect to BMI, since they have already been categorized
using the BMI thresholds. The mean BI between the 3 groups must be
different by definition.

With kind regards

Karabiner

#### obh

##### Well-Known Member
As to question 2) I am not sure what I might be missing:
2) See the differences between mean BMI within just the heart attack patients
It would make no sense to test whether normal/overweight/obese
differ with respect to BMI, since they have already been categorized
using the BMI thresholds. The mean BI between the 3 groups must be
different by definition.
Karabiner
This is what missing ... I assume Teddybear would want to check something else, not BMI ...

#### obh

##### Well-Known Member
Hi yes exactly. For 2 I would want to see if there is a difference between normal weight, overweight, and obese patients in just heart attack patients.
If you want to check if the mean BMI is different between (Heart attack patients, Neurological patients, and Other) groups you may run one one way ANOVA. H0: Mean(Heart attack patients) = Mean (Neurological patients) = Mean (Other)
If meets the test assumptions. (http://www.statskingdom.com/180Anova1way.html)