Use the Calculator to Support the Statistical Argument
Calculator technology can handle arithmetic quickly, but it does not decide what the question means, whether a procedure is appropriate, or what the result says in context. As in “Choosing the Correct Inference Procedure” and “Writing a Complete Inference Response,” begin with the variables, study design, and statistical question. Then choose the calculator function that matches that plan.
A calculator result is evidence to include in an explanation, not a replacement for one. For a probability question, identify the random variable and event, show the model and inputs, and interpret the probability. For inference, state the parameter and hypotheses or confidence level, name the procedure, check its conditions, and report the result and conclusion in context.
Match Probability Questions to Calculator Functions
For a normal model, use normalcdf(lower, upper, mean, standard deviation) to find the probability that the random variable falls between two values. Use a very large endpoint such as \(1\text{E}99\) to represent infinity, or a very small endpoint such as \(-1\text{E}99\) to represent negative infinity. For a cutoff associated with a specified area, use invNorm(area to the left, mean, standard deviation). The area input is cumulative area to the left, so a top-tail cutoff for 10% in the upper tail uses left area \(0.90\).
For a binomial random variable, binompdf(n, p, x) gives the probability of exactly \(x\) successes, while binomcdf(n, p, x) gives the probability of at most \(x\) successes. For example, \(P(X\geq k)\) can be found as \(1-P(X\leq k-1)\), so enter \(1-\text{binomcdf}(n,p,k-1)\). This complement step matters: the cumulative function includes its upper endpoint.
Worked Example: A Normal-Model Upper-Tail Probability
Scenario. Suppose the charging time \(X\), in minutes, for a certain model of device is described by a normal distribution with mean 42 minutes and standard deviation 6 minutes. Find the probability that a randomly selected device takes more than 50 minutes to charge.
Set up the event. The question asks for \(P(X>50)\). On a TI-84, enter normalcdf(50, 1E99, 42, 6), using 50 as the lower bound and a very large number as the upper bound. The calculator gives approximately \(0.09121\), or \(9.121\%\).
Check the result. Standardizing gives \(z=(50-42)/6=1.3333\). The standard normal upper-tail probability above \(z=1.3333\) is approximately \(0.09121\), consistent with the calculator result. In context, under this model, about \(9.121\%\) of devices take more than 50 minutes to charge.
What to report. State the event and the probability, then interpret it in context. Reporting only “normalcdf(50, 1E99, 42, 6)” does not tell the reader what was calculated or what the result means.
Worked Example: A Binomial Cumulative Probability
Scenario. A fictional seed packet contains seeds with a 0.50 probability of germinating, independently for each seed. Let \(X\) be the number that germinate in a group of 10 seeds. Find the probability that at most 3 germinate.
Choose the function. “At most 3” means \(P(X\leq 3)\), so use binomcdf(10, 0.50, 3). The result is \(0.171875\), or approximately \(0.1719\) rounded to four decimal places.
Interpret. Under the stated binomial model, the probability that no more than 3 of the 10 seeds germinate is about \(0.1719\), or \(17.19\%\). If instead the question asked for at least 8 germinating seeds, calculate \(1-\text{binomcdf}(10,0.50,7)=1-0.9453125=0.0546875\), approximately \(0.0547\). The subtraction uses 7 because the complement of \(X\geq 8\) is \(X\leq 7\).
Match Inference Questions to Calculator Procedures
The calculator’s inference menu offers several procedures, but procedure choice still follows the response type, number of samples, and study structure. Use the menu names as a way to carry out the plan from “Choosing the Correct Inference Procedure,” not as a substitute for that decision.
| Question structure | Common calculator procedure | What to report beyond the output |
|---|---|---|
| One population proportion | 1-PropZTest or 1-PropZInt | Define \(p\), enter \(x\) and \(n\), and report hypotheses or confidence level, conditions, and conclusion. |
| Two population proportions | 2-PropZTest or 2-PropZInt | Keep the groups in the stated order; identify \(p_1-p_2\) and interpret the direction of the difference. |
| One quantitative sample or paired differences | T-Test or TInterval | For paired data, enter or summarize the differences. Report the parameter as a mean or mean difference, as appropriate. |
| Two independent quantitative samples | 2-SampTTest or 2-SampTInt | Use the two groups in the stated order; for the test, use the unpooled option unless the problem specifically directs otherwise. |
| One categorical variable across groups, or two categorical variables in one sample | \(\chi^2\)-Test | Report the null and alternative in context, degrees of freedom, test statistic, p-value, and expected-count condition. |
For one- and two-proportion procedures, the calculator asks for counts and sample sizes, not just rounded sample proportions. For a \(t\) procedure, enter the data list or the summary statistics the menu requests, including sample size, mean, and standard deviation when using summary statistics. For paired \(t\) procedures, the input should represent the within-pair differences. For a chi-square test, enter the observed counts in a matrix; the calculator can produce expected counts, which you must still check.
Calculator menus can differ slightly by model. On a TI-84, the tests and intervals are available through the STAT TESTS menu. A test function typically reports a test statistic and p-value; an interval function reports endpoints. You may also use tcdf for a \(t\)-distribution probability and invT for a \(t\)-critical value, but a test or interval function is often less error-prone when its inputs match the problem.
Worked Example: A One-Proportion Test
Worked Example: Support for a Proposed Community Garden
State. A fictional town asks whether more than half of its adult residents support a proposed community garden. A random sample of 100 adults includes 60 who support it. Let \(p\) be the proportion of all adult residents of the town who support the proposal. The hypotheses are \(H_0:p=0.50\) and \(H_a:p>0.50\).
Plan and check conditions. Use a one-proportion \(z\) test. The sample is random. If the town has 1,800 adult residents, the 10% condition holds because \(100\leq 0.10(1800)=180\). Under the null hypothesis, the Large Counts condition holds: \(np_0=100(0.50)=50\geq10\) and \(n(1-p_0)=100(0.50)=50\geq10\).
Do. Enter 1-PropZTest with \(p_0=0.50\), \(x=60\), \(n=100\), and the alternative \(p>p_0\). The sample proportion is \(\hat p=60/100=0.60\). The test statistic is
The one-sided p-value is approximately \(0.0228\). This is the probability, assuming \(p=0.50\), of obtaining a sample proportion of \(0.60\) or higher from a random sample of 100 adults under the conditions of the model.
Conclude. Because \(0.0228\) is less than a significance level of \(0.05\), reject \(H_0\). The sample provides convincing evidence that more than half of the town’s adult residents support the proposed community garden. This conclusion is about the town’s adult residents represented by the random sample; it does not claim that every resident supports the proposal.
Worked Example: A One-Proportion Confidence Interval
Worked Example: A Library’s New Online Renewal Tool
Scenario. A fictional library randomly selects 120 cardholders from a population of 2,000 cardholders. Of those selected, 72 say they would use a new online renewal tool. Construct and interpret a 95% confidence interval for the proportion of all cardholders who would use it.
Plan and check conditions. Use a one-proportion \(z\) interval. Random selection is stated, and the 10% condition holds because \(120\leq0.10(2000)=200\). The sample has 72 successes and \(120-72=48\) failures, each at least 10.
Do. Enter 1-PropZInt with \(x=72\), \(n=120\), and confidence level \(0.95\). The sample proportion is \(\hat p=72/120=0.60\). The standard error used for the interval is \(\sqrt{0.60(0.40)/120}\approx0.04472\), and the 95% margin of error is approximately \(1.96(0.04472)=0.08765\). The resulting interval is approximately \((0.512,\ 0.688)\), rounded to three decimal places.
Interpret. We are 95% confident that the proportion of all cardholders in this library’s population who would use the new online renewal tool is between 0.512 and 0.688, or about 51.2% and 68.8%. The confidence level describes the long-run success rate of the interval method; it does not mean there is a 95% probability that this particular fixed population proportion lies in the calculated interval.
What to Write Down for Full Credit
A useful habit is to record the statistical setup before or alongside the calculator output. The display alone rarely includes everything an AP response needs. For inference, “Writing a Complete Inference Response” provides the four-part structure: State, Plan, Do, and Conclude. Calculator use fits inside the Do step; it does not replace the other steps.
Name the population parameter and the question. For a test, write \(H_0\) and \(H_a\). For an interval, state the parameter and confidence level.
Name the procedure and explicitly check randomization, independence or the 10% condition when relevant, and the appropriate Large Counts, shape, or expected-count condition.
Record the entered quantities, the test statistic and p-value, or the interval endpoints. Include a formula or enough substitution to show what the calculator evaluated.
For a test, say “reject” or “fail to reject” and describe the evidence in context. For an interval, interpret the endpoints for the population parameter in context.
A p-value is not the probability that the null hypothesis is true, nor is it the probability that the observed result happened “by chance” without qualification. It is calculated assuming the null hypothesis and the conditions of the test. An interval’s endpoints are also not meaningful by themselves: name the population parameter and its context.
Common Calculator Mistakes and AP Exam Tips
- Entering the wrong tail or endpoint. For “more than 50,” the lower bound is 50 and the upper bound is infinity. For “at least 8” in a binomial problem, use \(1-\text{binomcdf}(n,p,7)\), not \(1-\text{binomcdf}(n,p,8)\).
- Confusing a probability function with an inverse function. normalcdf returns an area; invNorm returns a value. Check whether the question asks “what is the probability?” or “what cutoff has this percentile?”
- Using the wrong counts or group order. Enter successes and sample sizes exactly as defined. In a two-group comparison, preserve the order in the question because reversing groups changes the sign and interpretation of the difference.
- Reporting only a command or screen output. A full-credit response names the method, shows relevant conditions and setup, and interprets the numerical result in context. A command is not a conclusion.
- Assuming the calculator checked conditions. It did not. Verify the 10% condition, Large Counts condition for a proportion \(z\) procedure, and the relevant conditions for \(t\) or chi-square methods yourself.
- Rounding too early or copying too few digits. Keep the calculator’s precision during intermediate steps, then round the final probability or endpoint consistently. Label a rounded answer and keep units or percentage points clear where needed.
- Overstating a test result. A small p-value can provide convincing evidence against \(H_0\), but it does not prove the alternative hypothesis or establish that an effect is large or practically important.
A reliable final check is to ask: Does the output answer the question I intended? Do the input order and tail match the wording? Have I stated the method and checked the conditions? Have I written a conclusion that says what the result means for the population or setting in the problem? Those checks turn calculator speed into clear statistical communication.
Check Your Understanding
For each question, choose a function or procedure and describe what you would report in addition to the calculator result.
- A normal random variable has mean 75 and standard deviation 8. Which calculator function would find the probability of a value between 70 and 82, and what four inputs are needed?
- A binomial count has \(n=12\) and \(p=0.30\). Write the calculator expression for the probability of at least 5 successes.
- A random sample of 80 adults includes 46 who favor a proposal. Which one-proportion test inputs are needed to test whether the population proportion is greater than 0.50? Name two checks you must make before using the result.
- What is the difference between the result from normalcdf and the result from invNorm?
- A calculator reports a p-value of 0.04 for a correctly planned test. Write one thing this p-value means and one thing it does not mean.