2
$\begingroup$

Let me give you an example to show what i am trying to. Let's say 50 patients started a cure for a disease and it took 2 months. So, I want to compare the first month's tumor sizes to the second month's tumor sizes for each of 50 patients. And i should be able to tell the patients tumor sizes are stable/increase/decrease over the cure period.

I have 50 patients' data that describe the effect of treatment in the first month and the second month. For example, the data looks like

patient-id  first-month  The size of tumor volume
1           1                 90
1           1                 89
1           1                 88
2           1                 90
2           1                 90
2           1                 91
1           2                 75
1           2                 80
1           2                 83
2           2                 98
2           2                 95
2           2                 97
..          ..                 ..
..          ..                 ..
50          2                 67

Here if the first month is 1, that means it is in the first month, 2 means it is in the second month of the treatment.

And the IDs are just number to represents each patient.

There are more than 3 measurements for each patient, but i have shown only 3 of them for brevity.

How to compare the first month of the treatment to the second month of the treatment for the disease. I used A one way Anova to compare the first month to the second month by using the "The size of tumor volume" as a dependent variable and "first-month" as a factor in SPSS. The results show there is a significant difference between the first month of the tumor size and second month of the cancer size for all patients. But, i did not include patient-ids in SPSS, should I use it? Should I compare each patient's first month data and second month of the treatment period separately? So, instead of having one Anova test result, should i have 50 (number of patients) Anova results to say that the differences is significant?

$\endgroup$
8
  • 1
    $\begingroup$ You should definitely be using the information in patient ID, since observations on a given patient are not independent of each other. But ID is not treated an independent variable; you use it to relate the measurements. If you had two months data, it would be paired, more generally you're dealing with repeated measures. $\endgroup$
    – Glen_b
    Jan 9, 2014 at 23:54
  • $\begingroup$ Yes, I know i should be using IDs, but in the Anova test i do not know how to include IDs in my test in SPSS. Do you know it? $\endgroup$
    – Co Koder
    Jan 10, 2014 at 0:26
  • $\begingroup$ As I already explained, you don't use ID as an independent variable - the fact that the individual patients have more than two measurements would lead you to use repeated measures. $\endgroup$
    – Glen_b
    Jan 10, 2014 at 0:53
  • $\begingroup$ So, the anove test that i got is ok? and make sense? $\endgroup$
    – Co Koder
    Jan 10, 2014 at 1:45
  • 1
    $\begingroup$ That question has already been covered by the first sentence of my first comment. As already explained, your ANOVA doesn't take account of the common patient IDs (e.g. doesn't use repeated measures), so no. $\endgroup$
    – Glen_b
    Jan 10, 2014 at 6:15

1 Answer 1

1
$\begingroup$

If I understand well what you are working with, I think you should apply repeated measures t-test. That is to compare all 50 patients' results before and after the treatment.

Note, that your data may not be normally distributed (as tumor volumes are known not to be).

As for the IDs you should specify what that stands for, because from the concrete example I cant see the sense.

$\endgroup$
6
  • $\begingroup$ Thanks for the reply. I do not want to compare all 50 patients' results before and after the treatment. Let me give you an example to show what i want. Let's say 50 patients started a cure and it took 2 months. So, I want to compare the first month's tumor sizes to the second month's tumor sizes for each of 50 patients. $\endgroup$
    – Co Koder
    Jan 9, 2014 at 22:53
  • $\begingroup$ And the IDs are just number to represents each patient. $\endgroup$
    – Co Koder
    Jan 9, 2014 at 22:55
  • $\begingroup$ What is the significance of the three measures within each individual within each month? Are spaced equally? The one-way ANOVA is problematic, because you are violating the assumption of independence. This seems like something that would benefit from a analysis that can properly model time, like a mixed-effects model $\endgroup$
    – dmartin
    Jan 9, 2014 at 23:47
  • $\begingroup$ @dmartin, you mean p value? the number of the measurements from the patients are not the same amount. One patient might have 30 measurements, other might 16. Why the assumption of independence is violated? So, if I use mixed effects model, what should be the fixed variables and random variables in SPSS? Or how to apply the mixed-effects model in SPSS? $\endgroup$
    – Co Koder
    Jan 10, 2014 at 0:35
  • $\begingroup$ By "significance" I didn't mean statistical significance, I was just wondering why splitting time by month was important to the research question, given that there were three time points per month. The independence assumption is violated in this dataset, because each individual is contributing multiple data points per month. Lastly, I can't really say what model might be appropriate until the three timepoints per month question is answered $\endgroup$
    – dmartin
    Jan 10, 2014 at 0:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.