2
$\begingroup$

In a crossover trial involving 50 patients, two dialysis treatment schedules were compared.Ten patients complained of vomiting on both schedules, 30 on neither, nine on treatment A only and one on treatment B.

(a) What are the proportions of patients vomiting on each treatment?

(b) Perform a significance test comparing these two proportions and interpret the results.

For (a), I think it is 19/50 and 11/50

For (b), I think I have to use McNemar test: (19-11)^2/(19+11)=2.31.

But I'm not sure, can you please check if my answer is right.

$\endgroup$

1 Answer 1

4
$\begingroup$

The way to check is to create the contingency table:
\begin{array}{rccl} &\text{treatment B} &\text{vomiting} & \\ \text{treatment A vomiting }\ \ &{\rm yes} &{\rm no} &{\rm sum} \\ {\rm yes} &10 &9 &? \\ {\rm no} &1 &30 &31 \\ {\rm sum} &? &39 & \end{array} An overview of McNemar's test, it's justification, and the formulas can be found here, and below that here. (To provide a hint, your formula is incorrect.)

$\endgroup$
4
  • $\begingroup$ Can you make it the rest of the way yourself, now? $\endgroup$ Commented Jun 21, 2016 at 20:00
  • 2
    $\begingroup$ To add a hint for the general problem of paired binary data: concordant results are effectively thrown away with McNemars... our attention is only on discordant pairs (those on the off-diagonal of the contingency table). $\endgroup$
    – AdamO
    Commented Jun 21, 2016 at 20:23
  • 1
    $\begingroup$ Ou, so the right formula then is (9-1)^2/(9+1)=6.4>>3.84, which means that we should reject the null hypothesis. Thank you! $\endgroup$
    – Art
    Commented Jun 21, 2016 at 20:57
  • 1
    $\begingroup$ You're welcome, @Art. Good luck with your course. $\endgroup$ Commented Jun 21, 2016 at 21:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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