Test of Significance for difference of Proportions

With this test we can test significance of difference between proportions of two
populations

Here null hypothesis is set as

𝐻0: 𝑃1 = 𝑃2 i.e. the difference between proportions is not significant

against the alternative hypothesis

𝐻1 : 𝑃1 ≠ 𝑃2 i.e. the difference between proportions is significant (Two-tailed

test)
Or 𝐻1: 𝑃1 > 𝑃2 (one tailed test, right tailed test)
Or 𝐻1: 𝑃1 < 𝑃2 (one tailed test, left tailed test)

Formulae:
1. If Population proportions 𝑃1 and 𝑃2 are known, then test statistic is
𝑝 −𝑝 𝑝 −𝑝
𝑧 = 𝑃 (1−𝑃1 ) 𝑃2 (1−𝑃 ) = 𝑃 𝑄1 𝑃2 𝑄
1 1 2 2 1 1 2 2
√ + √ 𝑛1
+ 𝑛
𝑛1 𝑛2 2
Where 𝑝1 = sample proportion of first sample
𝑝2 = sample proportion of of second sample
𝑛1 = sample size of first sample
𝑛2 = sample size of second sample
𝑃1 = Proportion of first population
𝑃2 = Proportion of second population
𝑄1 = 1 − 𝑃1
𝑄2 = 1 − 𝑃2

𝑃1 𝑄1 𝑃2 𝑄2
In this case standard error is 𝑆. 𝐸.𝑝1 −𝑝2 = √ +
𝑛1 𝑛2

2. If Population proportions 𝑃1 and 𝑃2 are unknown , then sample sample

proportions 𝑝1 and 𝑝2 can be used and test statistic is
𝑝1 −𝑝2
𝑧= 1 1
√𝑝𝑞(𝑛 +𝑛 )
1 2
Where 𝑝1 = sample proportion of first sample
𝑝2 = sample proportion of second sample
𝑛1 = sample size of first sample
𝑛2 = sample size of second sample
𝑛1 𝑝1 + 𝑛2 𝑝2 𝑋1 + 𝑋2
𝑝𝑜𝑜𝑙𝑒𝑑 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒 = 𝑝 = =
𝑛1 + 𝑛2 𝑛1 + 𝑛2
 𝑞 =1−𝑝

1 1
𝑆. 𝐸.𝑝1 −𝑝2 = √𝑝𝑞 ( + )
𝑛1 𝑛2
3. 95% confidence interval for difference between population proportions is
given by
𝑃1 (1−𝑃1 ) 𝑃2 (1−𝑃2 )
|𝑝1 − 𝑝2 | ± 1.96 𝑆. 𝐸.𝑝1 −𝑝2 = |𝑝1 − 𝑝2 | ± 1.96√ +
𝑛1 𝑛2
4. 99% confidence interval for difference between population proportions is
given by
𝑃1 (1−𝑃1 ) 𝑃2 (1−𝑃2 )
|𝑝1 − 𝑝2 | ± 2.58 𝑆. 𝐸.𝑝1 −𝑝2 = |𝑝1 − 𝑝2 | ± 2.58√ +
𝑛1 𝑛2
5. 100% confidence interval for difference between population proportions is
given by
𝑃1(1 − 𝑃1) 𝑃2(1 − 𝑃2)
|𝑝1 − 𝑝2 | ± 3 𝑆. 𝐸.𝑝1 −𝑝2 = |𝑝1 − 𝑝2 | ± 3√ +
𝑛1 𝑛2

 Before an increase in excise duty on tea , 400 people out of sample of 500
were found to be tea drinkers. After an increase in duty, 400 people were
tea drinkers in a sample of 600 people. Using standard error of proportion
, state whether there is significant decrease in consumption of tea at 5%
level of significance. Take value of Z at 5% level of significance as
1.645.
Sol. Here
Sample size of first sample = 𝑛1 = 500
Sample size of second sample = 𝑛2 = 600
No. of tea drinkers in first sample = 𝑋1 = 400
No. of tea drinkers in second sample = 𝑋2 = 400
𝑋 400
sample proportion of first sample = 𝑝1 = 1 = = 0.8
𝑛1 500
𝑋2 400
sample proportion of of second sample = 𝑝2 = = 0.67
𝑛2 600
Let 𝑃1 and 𝑃2 be population proportions of persons who are regular tea drinkers
before and after increase in duty respectively.
𝐻0 : 𝑃1 = 𝑃2 i.e. the difference between proportions is not significant and there
is no significant decrease in consumption of tea after increase in duty.

𝐻1 : 𝑃1 < 𝑃2 i.e. the difference between proportions is significant and there is

significant decrease in consumption of tea after increase in duty.

(One-tailed test)

𝑛1𝑝1 + 𝑛2 𝑝2 500(0.8) + 600(0.67)

𝑝𝑜𝑜𝑙𝑒𝑑 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒 = 𝑝 = =
𝑛1 + 𝑛2 500 + 600

400 + 400 800 8

= = = = 0.7272
1100 1100 11

𝑞 = 1 − 𝑝 = 1 − 0.7272 = 0.2728
𝑝1 −𝑝2 0.8−0.67 0.13
𝑧= = =
1 1 1 1 √(0.1984)(0.002+0.0016)
√𝑝𝑞(𝑛 +𝑛 ) √0.7272(0.2728)( + )
1 2 500 600
0.13 0.13 0.13
= = = = 4.8788
√(0.1984)(0.0036) √0.00071 0.026646

𝑦𝑖𝑒𝑙𝑑𝑠
→ |𝑧| = 4.8788
Critical value of z at 5 % level of significance for one tailed test is 1.645
Since |𝑧| = 4.8788 > 1.96, so we reject 𝐻0 & accept 𝐻1 and therefore there is
significant decrease in consumption of tea after increase in duty.

 A sample survey of tax payers belonging to Business class and Professional

class yielded the following results:
Business class Professional class
Sample size 400 420
Defaulters in tax 80 65
payment
Test the hypothesis that the defaulter rate is the same for the two classes of tax
payers.
Sol Here
Sample size of first sample = 𝑛1 = 400
Sample size of second sample = 𝑛2 = 420
No. of tax defaulters in first sample = 𝑋1 = 80
No. of tax defaulters in second sample = 𝑋2 = 65
 𝑋1 80
sample proportion of first sample = 𝑝1 = = = 0.2
𝑛1 400
𝑋2 65
sample proportion of of second sample = 𝑝2 = = = 0.1547
𝑛2 420

Let 𝑃1 and 𝑃2 be population proportions of tax defaulters in both the classes

respectively.
𝐻0 : 𝑃1 = 𝑃2 i.e. there is no significant difference in the defaulter rate for two
classes of tax payers.

𝐻1 : 𝑃1 ≠ 𝑃2 i.e. there is significant difference in the defaulter rate for two

classes of tax payers.

(Two-tailed test)

𝑛1𝑝1 + 𝑛2 𝑝2 80 + 65
𝑝𝑜𝑜𝑙𝑒𝑑 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒 = 𝑝 = =
𝑛1 + 𝑛2 400 + 420

145
= = 0.1768
820

𝑞 = 1 − 𝑝 = 1 − 0.1768 = 0.8232
𝑝1 −𝑝2 0.2−0.1547 0.0453
𝑧= = =
1 1 1 1 √(0.1455)(0.0025+0.0023)
√𝑝𝑞(𝑛 +𝑛 ) √0.1768(0.8232)( + )
1 2 400 420
0.0453 0.0453 0.0453
= = = = 1.7378
√(0.1455)(0.0048) √0.00069 0.026267

𝑦𝑖𝑒𝑙𝑑𝑠
→ |𝑧| = 1.7378
Critical value of z at 5 % level of significance for two tailed test is 1.96
Since |𝑧| = 1.7378 < 1.96, so we accept 𝐻0 i.e. there is no significant
difference in the defaulter rate for two classes of tax payers.

 In two large populations there are 30% and 25% fair haired people
respectively. Is this difference likely to be hidden in samples of 1200 and
900 respectively?
Sol. Here

Sample size of first sample = 𝑛1 = 1200

Sample size of second sample = 𝑛2 = 900
Proportion of first population = 𝑃1 = 30% = 0.3
Proportion of second population = 𝑃2 = 25% = 0.25
 𝑄1 = 1 − 𝑃1 = 1 − 0.3 = 0.7
𝑄2 = 1 − 𝑃2 = 1 − 0.25 = 0.75
Let 𝑝1 and 𝑝2 be sample proportions of fair haired people in both the samples
respectively.

𝐻0 : 𝑝1 = 𝑝2 i.e. the difference is likely to be hidden in samples.

𝐻1 : 𝑝1 ≠ 𝑝2 i.e. the difference is not likely to be hidden in samples.

𝑃1 − 𝑃2 𝑃1 − 𝑃2
𝑧= =
𝑃1(1 − 𝑃1 ) 𝑃2 (1 − 𝑃2) 𝑃1𝑄1 𝑃2𝑄2
√ + √ +
𝑛1 𝑛2 𝑛1 𝑛2
0.3 − 0.25 0.05 0.05
= =
√0.000175 + 0.000208
√0.3(0.7) + 0.25(0.75) √ 0.21 + 0.1875
1200 900 1200 900
0.05 0.05
= = = 2.5549
√0.000383 0.01957
Critical value of z at 5 % level of significance for one tailed test is 1.96
Since |𝑧| = 2.5549 > 1.96, so we reject 𝐻0 & accept 𝐻1 and conclude that the
difference is not likely to be hidden in samples.

NPTEL and Video Links

1. https://youtu.be/B7ybgf3CnJA (NPTEL link for difference of proportions)

2. https://www.khanacademy.org/math/ap-statistics/two-sample-inference/two-
sample-z-test-proportions/v/hypothesis-test-for-difference-in-proportions-example
3. https://www.khanacademy.org/math/ap-statistics/two-sample-inference/two-
sample-z-test-proportions/v/hypothesis-test-for-difference-in-proportions