SPSS Paired Samples T-Test outcome interpretation

Hypothesis:

-Blood pressure will decrease after intervention

Results (blood pressure):

Systolic bp before mean: 102.00

Systolic bp after mean: 97.10

Diastolic bp before mean: 70.80

Diastolic bp after mean : 65.90

----

Systolic bp (before and after) t: 1.884

Diastolic bp (before and after) t: 2.518

---

Systolic bp (before and after) 2-tailed sig.: .092

Diastolic bp (before and after) 2-tailed sig.: .033

Could you please help explain the significance of the results above and how I can use them to reject/retain the hypothesis?

I am also unsure on the level of significance to use: 0.05 or 0.1.

Thank you.

Your hypothesis is clearly directional, meaning that you do expect a decrease in blood pressure. Instead of just testing if there is a difference, you can test if there is a negative difference. This usually calls for 1-tailed testing. Nicely enough, you can just split the p-values in half for that (arriving at .046 and .0165).

Now it looks like both indicate a significant result, but: You tested two variables, which calls for some form of alpha-correction. The most conservative correction (Bonferroni) is to divide the alpha-threshold (let‘s use .05) by the number of tests, giving you .025. With this, you end up with only one significant result after all.

I assume that the two measures of blood pressure are correlated. You could consider using a multivariate test, which will test for differences in both dependent variables simultaneously. Unfortunately, interpretation of a significant multivariate test is tricky and you might end up doing t-tests anyway.

There are different ways of presenting significance. The „conventional“ way is to decide on a threshold beforehand (usually .05) and speak of significance based on that. There is no real reason, however, to use .05, which is way researchers often present the actual p-values, confidence intervals, etc. and treat significance with a little more... let‘s call it flexibility.

In any case, you should be aware of limitations regarding your sample and whether the differences are actually meaningful considering the variation in the data etc. Just because something is significant (or not) doesn’t mean the result is important (or isn’t). The absence of formal significance is also by no means a confirmation that there is no difference!

• Thank you ever so much for the explanation and elaboration. Yes, both measures are correlated. I am glad you pointed the need for alpha-correction. I will take your advice and use Bonferroni for correction. What would this mean in terms of rejecting/retaining the null hypothesis? Apr 18, 2019 at 11:45
• Sorry, I have another question; If I were to change my hypothesis to a non-directional so just looking to see if there is a difference, how would I go about doing that? Apr 18, 2019 at 11:46
• Take the p-values which you posted in the question (because they are already two-tailed) and correct for multiple tests by dividing the alpha threshold by the number of tests, e.g. .05 becomes .025. This would give you two non-significant results. To be clear, I do not recommend changing hypotheses after the fact, and I recommend trying a multivariate repeated measures design, which incorporates the correlation between the two dependent variables and tests them simultaneously. Apr 18, 2019 at 12:46
• Many thanks for the quick response. Just what I was looking for :) Apr 18, 2019 at 13:18

P-values should not be used as a definitive criterion for effectiveness of intervention--especially without knowing sample size or whether data are anywhere near normal.

Strictly according to usual 'rules for declaring significance' and taking t tests and associated the P-values to be valid, one would say that both systolic and diastolic are "signif" different at 10% level (2-sided paired t), and only diastolic at 5%.

It is wrong to discuss usefulness of intervention without a firm view whether a 5-point decrease in BP would be clinically meaningful (if indeed real).

• Sample consists of 10 female university students. Unfortunately, it's difficult to express on here but mean (Before) systolic bp ranged from 91 to 133; (After) systolic bp ranged from 87 to 115. (Before) diastolic bp from 60 to 80; (After) diastolic bp from 55 to 78. How would you suggest to calculate the effectiveness of intervention please? Thank you very much. Apr 18, 2019 at 11:32
• You can quite literally calculate effect sizes. Also, visualize your data. In your case, you could plot each value pair as a line, which will give you a nice overview of what is happening with individuals blood pressures. Apr 18, 2019 at 11:38