2.
Probability &
Statistics
BIOT F243 – Genetics
II Semester 2025-26
Aditya Sharma
aditya@[Link]
Probability Theory & Statistical Tests of hypotheses
Ø At the heart of Mendel’s work – luck vs insight?
Ø PT – Allows formulation of accurate predictions for experimental outcomes
Ø Stats – tests such as Chi-square allow testing of these hypotheses and attribute a
confidence score to interpretations based on the observed results
HOW TO EXPLAIN THE MISMATCH BETWEEN PREDICTION AND OBSERVATIONS?
Ø When dealing with Random / stochastic events – PT tells us WHAT TO EXPECT from data
Probability
Types of probabilities:
The probability (P) that an event will occur is the number of favorable cases (a) divided by
the total number of possible cases (n): P = a/n
Probability can be determined either by observation or by the nature of the event.
Examples:
Observation - One child in 10,000 is born with phenylketonuria.
Therefore, the probability of the next child born being born with phenylketonuria will be:
P = a/n = 1/10,000
Toss of dice`- the probability of any face to appear on tossing a dice is given by:
P = a/n = 1/6 (nature of the die is to have 6 faces)
Cards deck - the probability of drawing a seven of club from a deck:
P = a/n = 1/52
- The probability of drawing just a seven?
Probability
Mutually exclusive events, Independent events & Unordered events
Mutually exclusive events are events in which the occurrence of one possibility
excludes the occurrence of all others.
Eg. Tossing of a die – landing of any one face excludes the possibility of all other 5
Independent events are events whose outcomes do not influence one another.
Eg. Two dice tosses, the face value of one dice cannot influence the second dice.
Unordered events- are events whose probability of outcome does not depend on the
order in which the events occur.
Eg. The probability that a family of several children will have two boys, and one girl is the
same irrespective of the birth order. It does not matter whether the daughter is born first,
second or third.
Probability
Three rules for combining probabilities:
1. Sum Rule:
When events are mutually exclusive, the sum rule (either-or) is used
Eg. What is the probability of a dice showing either 4 or a 6 after being tossed in air
P = 1/6 + 1/6 = 2/6 = 1/3
2. Product Rule:
When the occurrence of one event is independent of the occurrence of the other events –
the probability of both the independent events occurring is the product of their
independent probabilities (and rule).
Eg. Throwing a dice two times and getting a four and then a six, in that order.
P = 1/6 x 1/6 = 1/36
3. Binomial Theorem:
It is used for unordered events. The probability that some arrangement will occur in which
the final order is not specified is defined by binomial theorem.
Eg. While tossing two coins simultaneously, what is the possibility of getting a head and
a tail?
Probability
Use of Rules:
There are several ways to calculate the probability for the above example.
1. By using a combination of rules 1 and 2
For each coin, the probability of getting a head (H) or a tail (T) is
For H: P = ½
For T: Q = ½
Tossing the coins one at a time, there are two ways of getting a head and a tail:
First head and then tail (HT), or First tail and then head (TH)
Within each ordered sequence (HT or TH), the individual outcomes are independent
events. Thus, the probability for any one of the two sequences involves the product
rule (rule 2):
½ x ½ = ¼ for HT or TH
The two sequences (HT and TH) are mutually exclusive. Thus, the probability of getting
either of the two sequences involves the sum rule (rule1):
¼ +¼ = ½
Thus, for unordered events, we can obtain the probability by combining rules 1 and 2.
Probability
The binomial theorem (rule 3) provide the shorthand method. To use rule 3, we must
state the theorem as follows:
If the probability of an event (X) is p and an alternative (Y) is q, then the probability
in n trials that event X will occur s times and Y will occur t times is:
In this equation, s+t = n and p + q = 1.
So, the probability of tossing two coins at once and getting one head and one tail:
In this case, n=2, s and t =1 and p and q = ½
Probability
What is the probability that a family with six children will have five girls and one boy?
Ø What is the probability that the family with six children will first have four girls, then one
boy, and then another girl?
Ø If the parents are heterozygous for albinism, and have 4 children:
a) what is the probability that all four children will be normal?
b) What is the probability that three will be normal and one albino?
c) What is the probability that the first three will be normal and the fourth one albino?
Probability
Ø The formula for rule 3 is the formula for the terms of the binomial expansion: If (p+q)n
is expanded the formula (n!/s!t!)psqt gives the probability for any one of these terms,
given that p+q =1 and s+t = n
Ø Since there are n+1 terms in the binomial, the formula gives the probability for the term
numbered (t+1)
Ø Since rule 3 is a binomial expansion formula, we can also generalize it to more than
two outcomes
Ø The general form for the multinomial expansion is ( p + q + r …) n and the general
formula for the probability is
Where s + t + u + ……… = n and p + q + r + ……. = 1
Probability
Ø If the parents heterozygous for albino have 5 children, what is the probability that they
will have two normal sons, two normal daughters and one albino son?
Statistics
Ø How much of a deviation is acceptable in observed vs expected values such that that
observation can be considered as reliable or accurate? How much approximation or
rounding up can be accepted?
Ø Any time we deal with probabilistic events there is some chance that the data will lead us
to support a bad hypothesis or reject a good one. Statistics quantifies these chances.
We cannot say with certainty that a 2.84:1 ratio represents a 3:1 ratio; we can say,
however, that we have a certain degree of confidence in the ratio. Statistics help us
ascertain confidence limits.
Ø In genetics, it can help us in the following ways:
Ø Setting the experimental design
Ø Summarizing the data – mean, median, standard deviation etc.
Ø Hypothesis testing
Testing of hypothesis
Ø Determining whether to support or reject a hypothesis by comparing the data to the
predictions of the hypothesis.
- was the ratio of 787:277 really indicative of a 3:1 ratio?
Statistics
Ø To begin, we need to establish how much variation to expect. This can be determined by calculating
a sampling distribution: the frequencies with which various possible events could occur in a
particular experiment.
Ø If we self-fertilized a heterozygous tall plant, we would expect a 3:1 ratio of tall to dwarf
plants among the progeny. If we looked at the first four offspring, what is the probability we
would see three tall and one dwarf plant?
Similarly, we can calculate the probability of getting all tall (81/256=0.32), two tall and two dwarf
(54/256=0.21), one tall and three dwarf (12/256=0.05), and all dwarf (1/256=0.004) in this first set of 4.
Statistics
Statistics
As sample sizes increase (from four to eight to forty), the sampling distribution takes on the
shape of a smooth curve with a peak at the true ratio of 3:1 (75% tall progeny)—that is, there
is a high probability of getting very close to the true ratio.
Statistics
Ø However, there is STILL a small chance that our ratio will be way off! Any ratio can arise in a
given experiment even though the true ratio is 3:1.
Ø So, at what point do we decide that an experimental result is not indicative of a 3:1 ratio?
Ø Statisticians have agreed on a convention - When all the frequencies are plotted as in the earlier
figure, we can treat the area under the curve as one unit and we can draw lines to mark 95% of this
area.
Ø Any r at i os i nc l uded w i t hi n 95% l i m i t s ar e c onsi der ed
supportive of the hypothesis of a 3:1 ratio – failing to reject
Ø Any ratio in the remaining 5% area is considered
unacceptable
Ø Similarly, rejection within 10% or 1% can also be considered
Ø Thus, it is possible to see whether the experimental data
support our hypothesis (in this case, the hypothesis of 3:1)
Ø One in twenty times (5%) we will make a type I error - we will
reject a true hypothesis. (A type II error is failing to reject a
false hypothesis)
Statistics
Ø To determine whether to reject a hypothesis, we must derive a frequency distribution for
each type of experiment
Ø Mendel could have used the distribution shown in figure 4.1 for seed coat or seed colour,
as long as he was expecting a 3:1 ratio and had a similar sample size
Ø What about independent assortment, which predicts a 9:3:3:1 ratio?
Ø A geneticist would have to calculate a new sampling distribution based on a 9:3:3:1 ratio
and a particular sample size
Ø Hence, statisticians have devised shortcut methods by using standardized probability
distributions - such as the t-distribution, binomial distribution, and chi-square distribution
Ø Each is useful for a particular kind of data; geneticists usually use the chi-square
distribution to test hypotheses regarding breeding data
Chi-square test
When the sample subjects are distributed among discrete categories such as tall, short,
yellow, green etc. chi-square distribution can be used to evaluate the data statistically
The formula for converting the experimental data to chi-square value is:
χ2 = Σ (O – E)2 / E
Where O = observed number for a category and E is the expected number for that category.
Σ signifies the sum of all categories.
Chi-square test
Ø These values however have no meaning on their own – hence we convert them to probabilities by
determining where the χ2 value falls in relation to the area under chi-square distribution curve
Ø To determine which of the two hypothesis is true, we refer to the chi-square table which contains
these already calculated probability values
Ø In tables 4.2 & 4.3 each category of data contributes to the chi-square value. Hence it will increase
with increase in number of categories.
Ø Degrees of freedom, the count of independent categories, helps keep track of the number of
categories
Ø Degrees of freedom = number of categories minus one.
Ø In this Mendelian experiment, degrees of freedom = 1 ( 2 categories tall and dwarf – 1; only one
is independent)
Ø In the table the left column shows the degrees of freedom; the numbers across the top of the table
are the probabilities.
Ø The first row depicts the chi square values for one degree of freedom at various probabilities.
Chi-square test
Ø We are interested in the values under the column 0.05.
Ø The probability is 0.05 of getting a chi-square value of 3.841 or larger by chance alone,
given that the hypothesis is correct.
Ø For Mendel’s experiment, the critical chi-square (at p = 0.05, one degree of freedom) is
3.841.
Ø This value can be compared to the to the calculated χ2 value (0.6 and 244.45). Since
the chi-square value for the 3:1 ratio is 0.6, which is less then the critical value of 3.841,
we do not reject the hypothesis of a 3:1 ratio.
Ø However, for 1:1 ratio the calculated value is far too large then the critical value and
hence the hypothesis is rejected.
Problem:
Mendel self fertilized pea plants with round and yellow peas. In the next generation he recovered the
following numbers of pea:
315 round and yellow peas
108 round and green peas
101 wrinkled and yellow peas
32 wrinkled and green peas
What is your hypothesis about the genetic control of the phenotype? Do the data support this hypothesis?
Hypothesis is a 9:3:3:1 ratio. The chi-square value for 3 degrees of freedom at p= 0.05 is 7.815
o e o-e (o-e)^2 ((o-e)^2)/e chi
315 RY 315 312.75 2.25 5.0625 0.01618705 0.470024
108 Ry 108 104.25 3.75 14.0625 0.134892086
101 rY 101 104.25 -3.25 10.5625 0.101318945
32 ry 32 34.75 -2.75 7.5625 0.217625899
556
Since the chi-square, 0.47 is less than the critical chi-square, we fail to reject our hypothesis
of two locus genetic control with dominant alleles at each locus.
Null hypothesis:
Ø Hypothesis testing, involves testing the assumption that there is no difference between the
observed and expected samples. Therefore, the hypothesis against which the data are
tested is referred to as the null hypothesis.
Ø If the null hypothesis is not rejected, then we say that the data are consistent with it, not
that the hypothesis has been proved. This is because supportive numbers could arise from
many reasons. Our failure to reject the hypothesis is tentative acceptance of the
hypothesis.
Ø The use of the 0.05 probability level as a cutoff for rejecting a hypothesis is a convention
called the level of significance, and the rejected data is said to be significantly away from
the expected data.
Ø 0.01 probability level is said to be highly significant.