two groups of patients with ovarian tumor divided according to their serum CA 19-9 le

[sameer]

two groups of patients with ovarian tumor divided according to their serum CA 19-9 levels.
Group I is with normal CA 19-9 and Group II is with elevated CA 19-9 levels.
ovarian tumor is bilateral in 20 cases in group I and it is bilateral in 35 cases in group 2.
now how do i calculate the p value for bilaterality in two groups????? which test should i use?

[PsychResearcherRASIS]

What is the hypothesis behind the test?

[sameer]

thanks for responding. i am suppose to see whether with elevated CA 199 causes bilaterality of the tumor or not...

[duskstar]

So you are testing for a difference between two proportions?

[sameer]

to find out statistical difference between the patient with normal CA199 and in patients with elevated CA 199

[sameer]

yes, its the difference between 2 proportions

[GretaGarbo]

I must say that this post confused me. Just like the others I thought that it would be natural to compare proportions, that is proportions with ovarian cancer among both healthy and unhealthy. That is, I wanted to have a control group with healthy patients where the serum CA was also measured.

Then we could calculate the proportion with ovarian cancer and compare the proportion between the two groups. Or we could, which would be almost the same thing, do a traditional chi-squared test among the four cells. (Two rows of healthy respectively un-healthy. And two columns with the two groups of serum CA levels.)

But the only information there is now is the number of cancer patients and no control group.

Let us think of the number of patients as a Poisson variable. Now we have observed one Poisson variable with the value of 20 and another one with the number of 35. These two groups have different serum CA levels. Is the Poisson intensity different between the two groups?

Of course this model is dependent on that there is a random sampling of the patient to the two groups. I guess that there could be many difficulties with this model. It would be nice to hear what the others here say about this.

Anyway, I tested this is R:

no.ovarian <- c(20, 35)
group      <- c(1,2)
group <- as.factor(group)

mod1 <- glm(no.ovarian ~ group, family=poisson  )
summary(mod1)

# Coefficients:
#             Estimate Std. Error z value Pr(>|z|)
# (Intercept)   2.9957     0.2236  13.397   <2e-16 ***
# group2        0.5596     0.2803   1.996   0.0459 *
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

So the result is that it is just significant, with at p-value of 0.046 jus below the 5% limit.

I should also say that I think it would be better if one had use the original serum CA level and not the grouped variable. Then the serum variable would be used as a regression “x-variable”. I think that would give better precision and lower p-values.

[sameer]

Laterality * GROUP199 Crosstabulation

Count
GROUP199
normal Ca 199 elevated Ca 199 total
Laterality unilateral 36 16 52
bilateral 8 3 11
total 44 19 63


now how do i calculate the p value in the bilateral group? to see whether elevated CA199 is statistically significant than the normal in bilateral group or not?