1. ## Is this possible?

Hi,

I have come stuck on a really simple question. I have access to excel and SPSS.

I have a dataset of response to a treatment. Sample size = 40. 30 responded and 10 did not. As it was a novel approach patients were not expected to respond ie. null = treatment does not lead to response.

Can I work out a p value from just this? If so how.

Sorry if it's very basic but would really value any help.

Thanks,

K

2. ## Re: Is this possible?

Is the following valid simple chi squared valid for this situation as follows?

Response
Observed N Expected N Residual
.00 10 20.0 -10.0
1.00 30 20.0 10.0
Total 40

Test Statistics
Response
Chi-Square 10.000a
df 1
Asymp. Sig. .002
a 0 cells (0.0%) have expected frequencies less than 5. The minimum expected cell frequency is 20.0.

Therefore p value 0.002 ?

I am a little confused should our expected value for "0 response" not be 40?

Thanks

3. ## Re: Is this possible?

You need to provide more detail to your question. 40 people, 30 had a change, 10 did not, no control group, correct?

4. ## The Following User Says Thank You to hlsmith For This Useful Post:

k093 (08-28-2014)

5. ## Re: Is this possible?

Thanks for the reply. Unfortunately there is no control group - not ideal I know but it just represents preliminary work.

6. ## Re: Is this possible?

Unless, you had a null hypothesis with a suspected % response, there really is no statistic for you. How did you define/quantify response, perhaps you can look at differences in values.

7. ## The Following User Says Thank You to hlsmith For This Useful Post:

k093 (08-28-2014)

8. ## Re: Is this possible?

I thought that may be the case. Response was defined as resolution of symptoms as reported by patient and/or a 20% drop in a blood test value. Would wilcoxon rank rum for change in blood test value between responders and non responders be reasonable ?

9. ## Re: Is this possible?

If you have those valules for all patients in both groups, you can calculate difference and use Wilcoxon sign test (analog to paired t-test).

10. ## Re: Is this possible?

Originally Posted by k093
I thought that may be the case. Response was defined as resolution of symptoms as reported by patient and/or a 20% drop in a blood test value. Would wilcoxon rank rum for change in blood test value between responders and non responders be reasonable ?
First you use blood test as a means to define responder and
non-responder, and the you look at whether responders and
non-responders differ with regard to blood test? Does not
sound too useful.

As to your first question: why do you think you need a p-value?
p-values ar means to an end, they help to decide whether
to maintain the Null hypothesis or reject the Null hypothesis.

Since you have at least 1 responder in your sample, the Null
hypothesis that in the population the response rate is exactely
= 0.00000% can safely be rejected.

As mentioned above by hlsmith, only some other expected
response rate (different from 0%) could be sensibly tested.

With kind regards

K.

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts