Signs and Evidence

Limitations

Multiple comparisons problem

Feeding multiple likelihood ratios into a calculation isn't helping the reliability of the outcome. This practice has a tendency to exaggerate the calculated effect of the signs used to reach that outcome. Because of the large number of variables, correcting this issue is impossible.

Sample size

Because (alas!) not all studies are freely made available online, it is not possible to select all available data for our database. This means that there will always be a bias due to the findability of diagnostic data. Paywalls and subscription fees are countless and prohibit an integral look at all data produced.

Weighing

Another reason is the fact that we treat all used studies as equal. Each individual study is of good quality, but naturally qualities differ. Meta-analyses are prefered over RCT's but their higher lever of proof is not weighed in calculating averages.

Not-weighing

Sometimes not all likelihood ratio’s are known, which might falsely imply that for some conditions the likelihood ratio is one. It is possible that the presence of a certain sign does not negatively or positively influence the calculated odds for a certain condition, due to lack of known statistical influencers.

Confidence intervals and probability values

Although only data with good p-values are included, confidence intervals can not be calculated based on the multifarious values from the different studies used. The numbers appearing as answers on your screen certainly have confidence intervals, but we have no way of estimating their size.