BOWEN UNIVERSITY IWO

COLLEGE OF AGRICULTURE ENGINEERING AND SCIENCE (COAES)

PURE AND APPLIED BIOLOGY PROGRAMME
BIO 201: GENETICS 1
Module 3
TEST OF GOODNESS OF FIT
Goodness of Fit
Goodness of fit refers to how well a statistical model fits a set of observations. In simpler terms,
it measures how close the predicted values from a model are to the actual observed values. It is a
crucial concept in statistical analysis because it helps assess the accuracy of models in
representing real-world data.
Goodness of fit is applicable in various types of models, including linear regression, logistic
regression, and other types of statistical models used in predictive analytics. Below are the key
components and methods used to evaluate goodness of fit.
Purpose of Goodness of Fit
The purpose of assessing the goodness of fit is to determine:
• How well the model explains the variation in the data.
• If the model assumptions hold.
• Whether the model can be used to make reliable predictions.
Goodness of fit helps to understand whether the relationship between variables that a model
predicts holds true across the sample.
Methods for Measuring Goodness of Fit
There are several statistical methods used to measure the goodness of fit depending on the type
of model and data being used:
a. R-squared (R²)
R-squared is a commonly used statistic to measure the goodness of fit in regression models. It
tells you the proportion of variance in the dependent variable that can be explained by the
independent variables in the model. An R² value of 0 means that the model does not explain any
of the variation in the response variable. An R² value of 1 indicates that the model perfectly
explains all the variation. A higher R² value suggests a better fit, though it does not guarantee that
the model is accurate.
b. Adjusted R-squared

1|O L AT U B I V I C TO R I A
Adjusted R-squared is a modification of R² that adjusts for the number of predictors in the
model. It accounts for the possibility that R² can increase simply by adding more variables, even
if they are not meaningful.
Adjusted R² gives a more accurate picture when comparing models with different numbers of
independent variables.
c. Chi-square Test
The chi-square goodness of fit test is used to compare observed data with data expected under a
specific hypothesis. It checks if the frequency distribution of a categorical variable matches
expected distributions.
Formula:
(O−E)2
χ2= ∑ 𝐸

Where O represents observed values and E represents expected values.

A small chi-square statistic means a good fit; a large chi-square statistic indicates a poor fit.
d. Residual Analysis
Residuals are the differences between observed values and model-predicted values. The
distribution of residuals provides insights into how well the model fits the data.
Types of Residuals:
o Standardized Residuals: Help detect outliers.
o Studentized Residuals: Consider the influence of each point on the model.
Ideally, residuals should have a random pattern when plotted against predicted values. Patterns
in residuals suggest a poor fit or that assumptions (e.g., linearity) have been violated.
e. Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC)
Both AIC and BIC are used to compare different models and assess goodness of fit, especially in
more complex models. AIC penalizes the number of parameters in the model but seeks the one
with the lowest value. BIC: Similar to AIC but applies a higher penalty for models with more
parameters. Lower AIC or BIC values indicate better-fitting models.
Evaluating the Model Fit
There are various ways to evaluate how well a model fits the data:
1. Visual Inspection
Scatterplots: In regression, scatterplots of observed vs. predicted values provide a clear picture of
how well the model performs. If points lie close to the line of perfect fit, the model is a good fit.

2|O L AT U B I V I C TO R I A
Residual Plots: These help in visualizing whether residuals are randomly distributed, indicating a
good fit.
2. Hypothesis Testing
Null Hypothesis: Often, when testing goodness of fit, the null hypothesis is that the model fits
the data well.
p-value: A low p-value (usually < 0.05) suggests that the model does not fit the data well, leading
to rejection of the null hypothesis.
Goodness of fit is a fundamental concept in statistical modeling that ensures that models
accurately reflect the data they are designed to represent. Various techniques like R-squared,
residual analysis, and chi-square tests offer ways to quantify how well a model fits. However,
understanding the context, assumptions, and limitations of each method is crucial in applying the
concept effectively.

Chi-squared test
This test was developed in 1900 by Karl Pearson (1857–1936), in part to investigate theories of
genetic inheritance.
Chi-Square is used to find out how the observed value of a given phenomenon is significantly
different from the expected value. In Chi Square goodness of fit test, the term goodness of fit is
used in order to compare the observed sample distribution with the expected probability
distribution. Chi-Square goodness of fit test determines how well theoretical distribution (such as
normal, binomial, or Poisson) fits the empirical distribution. In Chi-Square goodness of fit test,
sample data is divided into intervals. Then the numbers of points that fall into the interval are
compared, with the expected numbers of points in each interval. In regard to the procedure, set up
the hypothesis for Chi-Square goodness of fit test, that is set upo both null and alternative
hypothesis. a) Null hypothesis: In Chi-Square goodness of fit test, the null hypothesis assumes that
there is no significant difference between the observed and the expected value. b) Alternative
hypothesis: In Chi-Square goodness of fit test, the alternative hypothesis assumes that there is a
significant difference between the observed and the expected value. Calculate chi-square using the
formula and find out for the given degrees of freedom if chi-square value is significant at .05 or
.01 levels. If so reject the null hypothesis. If not significant accept the null hypothesis. 2.2.2 Testing
Hypothesis of Equal Probability The Chi-square test is a useful method of comparing
experimentally obtained results with those to be expected theoretically on some hypothesis. The
formula for calculating χ2 is
χ2 =∑[(O-E)2]/E
Where; O = observed frequency of a phenomenon or even which the experimenter is studying
E = expected frequency of the same phenomenon based on “no difference” or “null”
hypotheses.

3|O L AT U B I V I C TO R I A
 The use of the above formula can be illustrated by the following example.
Example 1:
An attitude scale designed to measure attitude toward co-education was administered on 240
students. They have to give their response in terms of favorable, neutral and unfavorable. Of the
members in the group 70 marked favorable, 50 neutral and 120 disagreed. Do these results indicate
significant difference in attitude?
The observed data is (O) given in the first row of table below
In the second row is the distribution of answer to be expected on the basis of null hypothesis (E),
if each answer is selected equally.
Table 1: Responses from subjects in regard to the attitudes
Calculations Favourable Neutral Unfavourable Total
O 70 50 120 240
E 80 80 80 240
(O-E) -10 -30 40
(O-E)2 100 900 1600
(O-E)2 / E 100 /80 900/80 1600/80
X 1.25 11.25 20 ∑(O- E)2 /E =
32.50

The formula of χ2 is χ2 = ∑(O- E)2 /E

χ2 = 1.25+11.25+20=32.5
d.f.=(r-1)(k-1)=2
Entering table of χ2 (which can be obtained from any book of statistic), we find in row d.f.=2 a
value 5.59 and 9.19 given under the heading .05 and .01 levels of significance respectively. Our
obtained value is 32.5, which is far above the given value in the table. Thus, our results will be
marked significant at .01 level. We discard the null hypothesis which stated that there will be no
difference in the attitude. But from the results it may be stated quite confidently that there is
difference in the attitudes of people towards coeducation. It may also be mentioned that these
results clearly indicate that the results are not due to any chance factor.
Example 2: A personnel manager is trying to determine whether absenteeism is greater on one
day of the week than on another. His record for the past year shows the following scores. Test
whether the absence is uniformly distributed over the week.
Table 2.
Calculation Monday Tuesday Wednesday Thursday Friday Total
O 23 18 24 17 18 100
E 20 20 20 20 20 100

4|O L AT U B I V I C TO R I A
 (O-E) 3 -2 4 -3 -2
(O-E)2 9 4 16 9 4

(O-E)2 / E 9/20=0.45 4/20=.20 16/20=.80 9/20=.45 4/20=.20

The formula of χ2 is
χ2 =∑[(O-E)2 /E] = 0.45+0.20+0.80+0.45+0.20 = 2.10
χ2 = 2.10
d.f.=(r-1)(c-1)
d.f.=(5-1)(2-1)

Critical value of χ2 at .05 level=9.488 (refer to statistical table given at the end of the statistics
book) Critical value of χ2 at .01 level=13.277 (refer to statistical table given at the end of the
statistics book) The computed value of χ2, i.e. 2.10 is less than the critical values at .05 and
.01significance levels, we conclude that χ2 is not significant and we retain the null hypothesis.
We can say that the deviation of observed absenteeism from expectation might be a matter of
chance. On the other hand, if the computed value of χ2 is more than 9.48 or 13.28, then the null
hypothesis is rejected and it may be concluded that there is significant difference in the
absenteeism that happens on different days of the week.
However, in our example we have found the chi-square value being lower than what is given in
the table and so we retain the null hypothesis stating that the absenteeism does not vary interms
of the days and that it is purely a chance factor.
Steps for Chi-square Testing
1) First set a null of hypotheses.
2) Collect the data and find out observed frequency.
3) Find out the expected frequencies by adding all the observed frequencies divided by number
of Categories (In Ist example 240/3=80, in II example 100/5=20).
4) Find out the difference between observed frequencies and expected frequencies. (O-E)
5) Find out the square of the difference between observed and expected Frequency. (O-E)2
6) Divide it by expected frequency. (O-E)2 /E. You will get a quotient
7) Find out the sum of these quotients.
8) Determine the degree of freedom and find out the critical value of χ2 from table.

5|O L AT U B I V I C TO R I A
9) Compare the calculated and table value of χ2 and use the following decision rule. Accept null
hypothesis if χ2 is less than critical value given in table. Reject the null hypothesis if calculated
value of χ2 is more than what is given in the table under .05 or .01 significance levels.

6|O L AT U B I V I C TO R I A