Help needed for COVID-19 stats

C

ChadiG

New Member
#1
#1
Hi guys, I really need your help in selecting which stats test to use. Here's my problem: I have a certain population, part of it was symptomatic for COVID-19 others no, PCR were done on all and it turned positive for some in both groups so a blood work out was done (serology) for the virus it also turned out positive for some and negativee for others.
My stats problem (finally) : Was the false negative (PCR+ serology -) was more elevated in the symptomatic or asymptomatic patients? is this significant?
Was the true positive (PCR+ serology + ) more elevated in the symptomatic or asymptomatic patients? is this significant?


Would Cochran-Mantel-Haenszel help?
Any help is welcome
 
noetsi

noetsi

Fortran must die
#2
#2
How is your dependent variable measured? Interval, ordinal....

If its interval scale I would just run linear regression.
 
hlsmith

hlsmith

Less is more. Stay pure. Stay poor.
#4
#4
Yes, simplistically if you have the three binary variables CMH should work. If you are looking at multiple outcomes you may think about correcting your alpha to be more stringent.

So every person had PCR and serology testing, right? My follow-up issue may be that you are likely not collecting PCR and serology at the same time period for everyone, index times differ. So a positive serology and negative PCR may not mean the same thing for each person, since they are at a different periods since exposure. This would mess a little bit with being able to generalize your results.
 
C

ChadiG

New Member
#5
#5
@hlsmith yes they are not collected at the same time. My goal is to find if the rate of false positive (PCR+ serology -) is higher for those who have symptoms and if the true positive (PCR+ serology +) is higher for those who have symptoms as compared to the people without symptoms and if these were statistically significant
Thanks for your help
 
hlsmith

hlsmith

Less is more. Stay pure. Stay poor.
#6
#6
I would watch out using the terms false positive and true positive since serology isn't perfect - but I get why you are using those terms. Are you going to define true positive rate out of all results, so true positives / (false positives + false negatives + true negatives)?
 
You must log in or register to reply here.