P(x)=∑x/total
P(A):        P(B):
P(Ā):        P(B̄):

P(x|y)=P(xy)/P(y)=xy/∑y
P(A|B):   P(A|B̄):           P(B|A):   P(B|Ā):
P(Ā|B):   P(Ā|B̄):           P(B̄|A):   P(B̄|Ā):
If P(A)=P(A|B) then A, B are independent; P(B)=P(B|A). Rows and columns are multiples of each other

P(x or y)=P(x)+P(y)-P(xy)
P(AorB): (TP+FN+FP)/total     P(AorB̄): (TP+FN+TN)/total
P(ĀorB): (TP+FP+TN)/total     P(ĀorB̄): (FP+FN+TN)/total
If A and B are disjoint, i.e. no overlap P(AB)=0, then P(A or B)=P(A)+P(B).

Accuracies:
Sensitivity= TP/(TP+FN): % of those actually with the disease who are found. TP/prevalence
   FNR (false negative rate)=1-sensitivity:
Specificity= TN/(TN+FP): % correctly those without disease. TN/(1-prevalence). FPR (false positive rate)
   FPR (false positive rate)=1-specificity:

PPV: positive predictive value: TP/(TP+FP)=P(A|B)
NPV: negative predictive value: TN/(TN+FN)=P(Ā|B̄)

Bayes factor likelihood ratio= sensitivity/FPR: P(+|disease)/P(+|no disease)