topic Using regression to determine a difference in MEANS in Statistical Procedures

Using regression to determine a difference in MEANS

IROCMath — Sun, 21 Jul 2019 22:19:15 GMT

Hi,

I'm new to SAS and am going through the basic programming modules. I have a data set and want to show a difference in the mean heights of men and women for only a specific age (age=20). I know the concept behind the model I need to use which is logistic. I'm not too familiar with SAS to be able produce what I know I need to do. My dataset contains many, many variables and observations, but I filtered the data to only run analysis for age=20. I am working with GENDER (1=male, 2=female), LENGTH (which is height), and AGE (only needing to look at 20).

The response is LENTH vs the GENDER of the individuals for a specific AGE. Any suggestions for commands that may help here...

I cooked up this on SAS (University Edition):

proc glm data=hw2.z_final;
	class gender;
	where age=24;
	model length=gender/ effectsize alpha=0.05 Solution;
	run;

Thanks in advanced. If I need to clarify, let me know.

Re: Using regression to determine a difference in MEANS

PaigeMiller — Sun, 21 Jul 2019 23:56:09 GMT

This is not a Logistic Regression, it is probably a t-test, which can be performed in either PROC TTEST or PROC GLM (they should give the same answers).

Other than that, it seems like you have done the right thing.

Re: Using regression to determine a difference in MEANS

Reeza — Mon, 22 Jul 2019 01:48:00 GMT

Continuous outcome => GLM, ANOVA, T-TEST

Categorical or Binary outcome => LOGISTIC

https://www.tutorialspoint.com/sas/sas_ttests

GLM treatment of binary variables can be non-intuitive (check the design matrix in your output) so I recommend PROC ANOVA or TTEST instead.

proc anova data=sashelp.cars;
where make in ("Acura", "BMW");
class make;
model mpg_city = make;
run;

@IROCMath wrote:

Hi,

I'm new to SAS and am going through the basic programming modules. I have a data set and want to show a difference in the mean heights of men and women for only a specific age (age=20). I know the concept behind the model I need to use which is logistic. I'm not too familiar with SAS to be able produce what I know I need to do. My dataset contains many, many variables and observations, but I filtered the data to only run analysis for age=20. I am working with GENDER (1=male, 2=female), LENGTH (which is height), and AGE (only needing to look at 20).

The response is LENTH vs the GENDER of the individuals for a specific AGE. Any suggestions for commands that may help here...

I cooked up this on SAS (University Edition):
proc glm data=hw2.z_final;
	class gender;
	where age=24;
	model length=gender/ effectsize alpha=0.05 Solution;
	run;
Thanks in advanced. If I need to clarify, let me know.

Re: Using regression to determine a difference in MEANS

IROCMath — Mon, 22 Jul 2019 04:48:54 GMT

hmmm... why did my prof say it was log reg...? Ok, thanks!

Re: Using regression to determine a difference in MEANS

PaigeMiller — Mon, 22 Jul 2019 10:48:45 GMT

@IROCMath wrote:

hmmm... why did my prof say it was log reg...?

You could ask your prof to explain.

Re: Using regression to determine a difference in MEANS

PaigeMiller — Mon, 22 Jul 2019 12:40:33 GMT

@Reeza wrote:

Continuous outcome => GLM, ANOVA, T-TEST

Categorical or Binary outcome => LOGISTIC

https://www.tutorialspoint.com/sas/sas_ttests

GLM treatment of binary variables can be non-intuitive (check the design matrix in your output) so I recommend PROC ANOVA or TTEST instead.

My usual warning:

Although PROC ANOVA will work here in this specific example, it should not be used for most problems of this type. So sayeth the documentation:

Use PROC ANOVA for the analysis of balanced data only, with the following exceptions: one-way analysis of variance, Latin square designs, certain partially balanced incomplete block designs, completely nested (hierarchical) designs, and designs with cell frequencies that are proportional to each other and are also proportional to the background population. These exceptions have designs in which the factors are all orthogonal to each other.

Re: Using regression to determine a difference in MEANS

Reeza — Mon, 22 Jul 2019 14:52:24 GMT

Are you sure you're interpreting the problem correctly?

@IROCMath wrote:

hmmm... why did my prof say it was log reg...? Ok, thanks!