Dear Stats people, I desperately need your help.
I have created a laser to diagnose Covid-19 but have to supply Stat data for ethics and medical device approval.
May I have your opinion please - because, I know nothing about this field of maths and logic.
In Australia ~250000 tests with 2.14% positive PCR tests. I am hoping to validate the positive side of tests as a negative PCR test does not mean its negative - it just means its negative or you did not acquire the RNA. There is an expectation there may be a 5 to 10 fold greater infection rate than actually tested, partly to not testing and partly to false-negative tests.
So I am assuming in my validation trial I will at least get a doubling of the 2.14% positive rate to 4.28%. Its a guess but I have to choose a figure.
I have used this calculator:
https://clincalc.com/stats/samplesize.aspx
Sorry the formula does not paste well
Incidence, population 2.14%
Incidence, study group 4.28%
Alpha 0.05
Beta 0.05
Power 0.95
=p0q0{z1−α/2+z1−βp1q1p0q0−−−√}2(p1−p0)2q0=1−p0q1=1−p1N=0.0214∗0.9786{1.96+1.640.0428∗0.95720.0214∗0.9786−−−−−−−−−√}2(0.0428−0.0214)2
N=830
p0 = proportion (incidence) of population
p1 = proportion (incidence) of study group
N = sample size for study group
α = probability of type I error (usually 0.05)
β = probability of type II error (usually 0.2)
z = critical Z value for a given α or β
So the number of tests is at least N =830 . I was planning on at least 1000
people in the validation trial.
So I have some questions - in Stats - how do you handle this data/present this data? Sorry, I am a complete novice in this field.
eg these are the questions I am being asked :
• Data Analysis: How will you measure, manipulate and/or analyze the information that you collect/gather?
o Matching and sampling strategies
o Accounting for potential bias, confounding factors and missing information
o Statistical power calculation
Thanking you in advance