Hi all,
So I have several triage pathways using different tests. The sample cohort is the same however and all data is categorical (positive or negative). Is using the Chi Squared in this instance applicable to see if there is a difference between each of the test on the population?
McNemars is another option but I dont think it is applicable in this instance.
Thanks in advance
Stephen
My Dear
You can use to test statistics the first Chi-square test statistic to test independent between variables study the second one is Z-test statistic to test proportion of one proportion or two
with Best wishes
Hello Stephen,
It depends on your research aim. If you're trying to determine whether alternate tests lead to the same result, then it's the agreement (instances of +/+ or -/-) that is of interest. Cohen's coefficient kappa quantifies the degree of agreement beyond chance level (which is determined by the marginal proportions), and is a very useful statistic for this sort of thing.
On the other hand, if you're more interested in whether there is a significance discrepancy between the classifications of alternate tests, then a test like McNemar's would be better suited. It evaluates only the disagreement categories (+/- or -/+ cells). These would be useful if you were looking for differential sensitivity/specificity.
Here's a link to a comparable discussion from years ago: https://www.researchgate.net/post/How_can_we_compare_sensitivity_of_two_test_to_prove_that_one_is_better_than_other
Good luck with your work!
Thanks David Morse and Fadhil Abdulabbas Abidi .
The aim is to show if there is a difference between test 1 vs test 2 in the study cohort. Same population, different test.
If you are only interested in testing the probability of a difference between the two tests in the population then use the Chi Square test. However, if you have a very large sample and there is even a small difference between the two tests you will get a significance. Similarly, if you have a very small sample you may fail to get significance when, in fact, there is a difference. A power analysis will give you an estimate of the sample size you need for a given effect size. There are a few power calculators available on the internet.
Use independent chi-square test if the groups are independent to each other (e.g. Male and female) however if the data is paired like Pre and Post (e.g. The response of the patient (YES/NO) about any disease is noted at Baseline, when the patients came first time in the hospital after one month follow up the same patient give his response about the same disease(YES/NO) so this type of data is paired in category therefore McNemars test is suitable for this data.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics