Statistical test between two variables?

[Justin12]

drug	placebo
105	160
110	140
108	150
112
	136
116	146
112	140
108
	150
114	138
	148
	150
	156
	152

I have this data that compares blood pressures. The placebo column is patients normal blood pressures and the drug column has their pressures after the drug is administered. I am trying to use a certain PROC statement to determine statistically whether the drug does lower pressures. So after I imported the data into SAS named 'blood' and I am running a simple t-test.

proc ttest data=blood;

paired drug*placebo;

run;

Does this make sense or should I use proq freq with the chisq option?

[ballardw]

If you research question is about "mean difference of measurement" then you ttest looks right though I hope you have more records.

Do take note than any subject that does not have both measurements will be excluded from the calculations.

If the same subject appears multiple times then you are into a "repeated measures" environment and ttest may not be appropriate.

[Reeza]

For a test like this you should run a paired test ideally, but do you have enough data? The documentation for PROC TTEST has an example I believe.

---

@Justin12 wrote:

drug	placebo
105	160
110	140
108	150
112
	136
116	146
112	140
108
	150
114	138
	148
	150
	156
	152

I have this data that compares blood pressures. The placebo column is patients normal blood pressures and the drug column has their pressures after the drug is administered. I am trying to use a certain PROC statement to determine statistically whether the drug does lower pressures. So after I imported the data into SAS named 'blood' and I am running a simple t-test.

proc ttest data=blood;

paired drug*placebo;

run;

Does this make sense or should I use proq freq with the chisq option?

---

[Justin12]

Ok, thanks. I do have more data, the above was just a small sample of it.

[Reeza]

Then a T-Test is the appropriate statistical test in this case, and a paired t-test. You should always make a graph too, a slope graph for all users and a histogram of the differences would be my suggestions.