The four counts add to the total, which is worth checking before you start:
70+20+8+2=100✓
(i) Type O
P(O)=10070=107
(ii) Type A or Type B
A sample cannot be both Type A and Type B, so no sample is counted twice and the two counts simply add:
20+8=28
P(A or B)=10028=257
(iii) Not Type AB
P(AB)=1002=501
P(not AB)=1−501=5049
(iv) Sent for cross-matching
"Not Type O" is the complement of Type O, so there is no need to add the other three counts:
P(not O)=1−107=103
(Adding directly gives the same thing: 20+8+2=30, and 10030=103.)
(v) After 20 more Type AB samples are added
Again both parts of the fraction change — update each in turn.
Total samples:
100+20=120
Type AB samples:
2+20=22
P(AB)=12022=6011
The point of parts (iii) and (v) together
Type AB went from 501=0.02 to 6011≈0.183 — roughly nine times as likely.
Adding samples of one type raises that type's probability and lowers every other type's, because the denominator grows for all of them. Type O, for instance, falls from 10070=0.70 to 12070≈0.583, even though not a single Type O sample was removed.