AP Statistics Topic 4.9: Setting Up a Test for the Difference Between Two Population Means

01 · Start here

The concept in 60 seconds

A two-sample $t$ -test asks whether data from $2$ independent groups provide convincing evidence of a difference between their population means. The response must be quantitative, such as time, height or score.

The setup has three jobs: Name the population means, state the competing claims about their difference, and check whether the study supports the test.

The null hypothesis is the starting benchmark: no difference in the population means. The alternative states the difference you are looking for: greater than, less than, or different from $0$ .

This lesson builds that plan. A sample mean difference alone does not tell you whether the population means differ. The test statistic, $p$ -value and final decision come in Topic 4.10.

02 · One situation, deeper understanding

Does service A take longer on average?

A delivery company asks whether service A’s population mean delivery time is longer than service B’s. It independently selects an SRS of $40$ of service A’s $2{,}000$ deliveries and an SRS of $36$ of service B’s $1{,}800$ deliveries from a defined period.

Delivery sample summaries
Service	Sample size	Sample mean	Sample $\operatorname{SD}$
A	$40$	$32 \,\mathrm{min}$	$8 \,\mathrm{min}$
B	$36$	$28 \,\mathrm{min}$	$6 \,\mathrm{min}$

Let $\mu_{A}$ and $\mu_{B}$ be the population mean delivery times, in minutes, for these $2$ services during that period. Keep the order $\mathrm A-\mathrm B$ .

$H_{0}: \mu_{A}-\mu_{B} = 0$
$H_{a}: \mu_{A}-\mu_{B} \gt 0$

The alternative is positive because the original question asks whether A takes longer. The observed difference, $32-28=4 \,\mathrm{min}$ , is a sample result; it is not the value to put in the alternative hypothesis.

The samples are independently selected, each is at most $10\%$ of its delivery population, and both sample sizes are at least $30$ . These facts support an independent two-sample $t$ -test. We have a valid setup, but have not yet calculated the evidence against the null.

03 · Understand the language

Key ideas and notation

Population means: $\mu_{1}$ and $\mu_{2}$

The unknown average responses for the $2$ defined populations. Specify who or what belongs to each population, the response and its units.

$\mu_{A}$ is the population mean time for service A’s deliveries during the study period.

Mean difference: $\Delta$

$\Delta =\mu_{1}-\mu_{2}$ is the population parameter being tested. Positive and negative values depend on which group comes first.

$\Delta =0$ means equal population means; it does not mean identical individuals or distributions.

Sample means: $\bar x_{1}$ and $\bar x_{2}$

These are calculated from the observed data. Their difference estimates $\Delta$ and later helps measure evidence against $H_{0}$ .

Hypotheses concern $\mu_{1}$ and $\mu_{2}$ , not the already observed $\bar x_{1}$ and $\bar x_{2}$ .

Null and alternative

$H_{0}$ states $0$ population mean difference. $H_{a}$ states the direction or difference the investigation seeks.

Choose $H_{a}$ before inspecting the sample outcome.

Independent two-sample $t$ -test

A test for a difference between $2$ population means using $2$ independent groups and sample standard deviations because population standard deviations are unknown. The unpooled method allows the population variances to differ.

04 · Match the method to the design

Choose the method before the hypotheses

Start with the response and how the observations were collected. Having $2$ columns of numbers does not automatically mean you have $2$ independent samples.

Which means procedure fits the design?

1

groupOne-sample $t$ -test

Compare $1$ population mean with a stated value. Example: mean package weight versus $500 \,\mathrm{g}$ .

Matched pairsOne-sample $t$ -test on differences

The same people measured $2$ times, or deliberately matched pairs. Compute $1$ difference per pair.

2

independent groupsTwo-sample $t$ -test

Separate groups without within-pair matching. Compare their population mean quantitative responses.

Categorical responseA proportions procedure

Success/failure or yes/no outcomes concern proportions. A test for means does not fit that response.

After choosing a method, check its conditions. This diagram identifies the design; it does not guarantee that inference is valid.

Test or interval?

Use a confidence interval to estimate a population mean difference. Use a hypothesis test to assess evidence against a specified claim. “How much higher?” calls for estimation; “Is there convincing evidence it is higher?” calls for testing.

Why $t$ , and why unpooled?

In these comparisons, the population standard deviations are unknown and the sample standard deviations estimate them. The $t$ procedure accounts for that estimation. The usual unpooled two-sample method does not assume equal population variances. Equal-looking sample standard deviations are not a reason to silently switch to a pooled procedure.

05 · Translate the claims

Write the null and alternative hypotheses

First define the parameters in a sentence. Then write both hypotheses using the same subtraction order.

$H_{0}: \mu_{1}-\mu_{2} = 0$
Equivalent form: $H_{0}: \mu_{1} = \mu_{2}$

The null includes equality. Choose one alternative from the original question:

Translate the question using group 1 minus group 2
Investigative question	Alternative	Type
Is group 1’s population mean higher?	$\mu_{1}-\mu_{2} > 0$	One-sided
Is group 1’s population mean lower?	$\mu_{1}-\mu_{2} < 0$	One-sided
Are the population means different?	$\mu_{1}-\mu_{2} \ne 0$	Two-sided

A parameter definition that earns its place

“ $\mu_{1}$ is group 1” is incomplete. Write: “ $\mu_{1}$ is the population mean task-completion time, in seconds, for users of the new interface.” Define $\mu_{2}$ just as precisely for the old interface.

Check the response: For time, “faster” usually means a lower mean. For score, “better performance” may mean a higher mean. Translate the meaning of the measurement, not just a positive-sounding word.

Do not write $H_{a}: \mu_{1}-\mu_{2} = 4$ merely because the observed sample difference is $4$ . A directional claim covers a range of possible population differences, and the population value remains unknown.

06 · Keep the direction consistent

Choose the direction from the question

A one-sided alternative looks for a difference in $1$ specified direction. A two-sided alternative looks for a difference in either direction. The study question decides this choice before the results are examined.

Read the alternative on a population-difference number line

$\Delta =\mu_{1}-\mu_{2}$ . These highlighted number-line regions show parameter values allowed by each alternative. They are not $p$ -values, probability distributions or rejection regions. The open circle excludes $0$ .

Reversing the subtraction order

If the claim is that A’s mean time is longer, you may write $\mu_{A}-\mu_{B} > 0$ or $\mu_{B}-\mu_{A} < 0$ . Both describe the same claim. Define your order and keep it consistent throughout the test.

Reversing the order changes the signs, including the sign of the later test statistic. It does not change the evidence for the same substantive claim when the alternative is reversed consistently.

Think before revealing: A study asks whether $2$ brands have different mean battery lives. The first sample mean happens to be larger. Should the alternative become “greater than”?

Reveal the answer

No. “Different” calls for a two-sided alternative, $\mu_{1}-\mu_{2} \ne 0$ . Changing to a one-sided test after seeing the result changes the question to favor the observed direction.

07 · Justify the plan with evidence

Check the conditions separately

Do more than name a checklist. Connect each condition to a fact about the actual study. Check the $2$ groups separately where appropriate.

1. Randomization and independent groups

Use $2$ independently selected random samples or a properly conducted randomized experiment assigning individuals to $2$ independent treatment groups. Repeated measurements on the same people are paired and require a different analysis.

Random sampling supports generalization to the sampled populations. Random assignment supports a causal comparison of treatments. The two serve different purposes. A large convenience sample does not become random simply because it has many observations.

2. The $10\%$ condition when sampling without replacement

$n_{1}\le 0.10N_{1}$ and $n_{2}\le 0.10N_{2}$

Each sample should be at most $10\%$ of its own population to support treating the sampled observations as approximately independent within that group. For the delivery study: $40\le 200$ and $36\le 180$ .

This sampling-without-replacement check is not required merely because a randomized experiment assigns participants to treatments. Do not invent a population size for a random-assignment problem. If a study also uses random sampling, assess that sampling step where relevant.

3. Sample size or distribution shape

The sample-size condition is met if both $n_{1}\ge 30$ and $n_{2}\ge 30$ . Alternatively, both populations may be known to be approximately normal.

If either sample has fewer than $30$ observations and population normality is not established, inspect both sample distributions: neither should show strong skewness or outliers. A small sample needs convincing shape evidence; do not claim that an unknown population is normal merely because $n$ is small.

Small samples: look at both groups

      EXCELLENT  
   Based on 51 reviews 
    
 
 
 
   
 
 
       Posted on Google  
   
   0000003998 : Abdad Alam Shamim Alam 
 
 
 
 Trustindex verifies that the original source of the review is Google. This is the bestest institution I have ever come to and I love it very much.
  
 
      Posted on Google  
   
   Aliki S 
 
 
 
 Trustindex verifies that the original source of the review is Google. 
  
 
      Posted on Google  
   
   Ali 
 
 
 
 Trustindex verifies that the original source of the review is Google. Sly is him
  
 
      Posted on Google  
   
   Jitendra Kumar Kumawat 
 
 
 
 Trustindex verifies that the original source of the review is Google. 
  
 
      Posted on Google  
   
   Raahil Hasan 
 
 
 
 Trustindex verifies that the original source of the review is Google. 
  
 
      Posted on Google  
   
   Tina Mudarres 
 
 
 
 Trustindex verifies that the original source of the review is Google. The classes are amazing and my child learnt so much
  
 
      Posted on Google  
   
   alisha gadoya 
 
 
 
 Trustindex verifies that the original source of the review is Google. I have experienced a lot of good things, it has taught me so many things and from B/Cs I have gone to an A! This is wonderful and Mavish’s class is awesome.
  
 
      Posted on Google  
   
   smasher 123 
 
 
 
 Trustindex verifies that the original source of the review is Google. It’s sensational my kid went. First he only would get C now it’s all A’s really good and recommend
  
 
      Posted on Google  
   
   Mosa Al- Samaraie 
 
 
 
 Trustindex verifies that the original source of the review is Google. AMAZING PRICES GREAT TEACHERS STRAIGHT FORWARD LEARNING STEADY PACE IN TUTORING
  
 
      Posted on Google  
   
   Cael Dagnelie 
 
 
 
 Trustindex verifies that the original source of the review is Google. I had a great experience learning math with Ms. Mavish. She explains complex topics in a very clear and simple way, which made it easier for me to understand and enjoy the subject. Her patience and dedication really stood out, and she always made sure that everyone in the class was keeping up. I especially appreciated how approachable she was — I never felt afraid to ask questions, and she was always willing to help. Thanks to her teaching, my confidence in math has grown a lot. I’m really thankful for the effort she puts into every lesson!

NUM8ERS is one of finest tutoring institutes in UAE, Located in Al Barsha 1, Dubai. Close to DUBAI AMERICAN ACADEMY (DAA) & AMERICAN SCHOOL OF DUBAI (ASD).

Setting Up a Test for the Difference Between Two Population Means

By the end of this lesson, you should be able to:

The concept in 60 seconds

Does service A take longer on average?

Key ideas and notation