HSA 523 Health Data Analysis Dr. Robert Jantzen Answer key for Homework 7 NOTE: click here for an MS-Excel spreadsheet that will conduct the following one sample tests: the z test for a population proportion and the t test for a population mean. 1. The proportion of short-term acute hospitals in the US that are investor owned for-profits is .18, or 18%. A much smaller number (.098 or 9.8%) of the 235 hospitals sampled by Loubeau and Jantzen were for-profits. A. What are the data requirements for conducting a z test for a population proportion? The data requirements include a random sample and n*ps >= 5 & n*(1-ps) >=5. B. Is the Jantzen/Loubeau sample large enough to test hypotheses about the population proportion? Yes. 23 of the hospitals were for profit (>=5) and 212 were non profits (>=5). C. Test whether the sample could have been drawn from a population of hospitals that had an investor owned proportion of .18 (18%). Show all work, including the hypotheses, sample statistic, critical statistic and decision rule. What can you conclude about the representativeness of the sample? This is a z test for the population proportion. Ho: population proportion = .18 Ha: population proportion not equal .18 Sample z (from calculator) = -3.28 and Critical z = 1.96 (two-tailed value) Since the sample z (in absolute value) is >= critical z, we reject the Ho. There is sufficient evidence that the population proportion that this sample was drawn from is not .18 meaning that this sample is not representative of the population of hospitals. D. What does the estimated p value (also called the sig. level) of the test show? The p value is .001 meaning that there’s only a .001 probability (.1% chance) of finding a sample proportion that differs from the null hypothesis value as much as this sample does, when the null hypothesis is true (that’s why we reject the Ho). It’s the probability of rejecting the null when it’s true. 2. Assume that JCAHO regulations require that at least 90% of a hospital's technical staff must meet current continuing educational requirements. Also assume that a random sample of 40 employees finds that 6 have not met those requirements. A. Is there sufficient data for conducting a z test for a population proportion. Yes. A random sample is required and that there are at least 5 employees who meet the requirement and 5 who don’t. B. Test whether the JCAHO regulation is being violated, i.e., that more than 10% of all employees don't meet the requirement. Show all work, including the hypotheses, sample statistic, critical statistic and decision rule. This is a z test for the population proportion. Ho: population proportion <= .10 Ha: population proportion > .10 Sample z (from calculator) = 1.05 and Critical z = 1.65 (one-tailed value) Since the sample z (in absolute value) is < critical z, we cannot reject the Ho. There is insufficient evidence to state that the population proportion in violation is > .1. C. What does the estimated p value (also called the sig. level) of the test show? The p value is .146 meaning that there’s a .146 probability (14.6% chance) of finding a sample proportion that differs from the null hypothesis value as much as this sample does, when the null hypothesis is true (that’s why we can’t reject the Ho). It’s the probability of rejecting the null when it’s true. 3. A manufacturer of radon detectors placed a random sample of twelve detectors in a chamber that exposed them to 105 picocuries per liter of radon. The detector readings were as follows: 91.9, 97.8, 111.4, 122.3, 105.4, 95.0, 103.8, 99.6, 96.6, 119.3, 104.8, 101.7 A. Create a data file using SPSS containing the above detector readings and find the mean and standard deviation for the sample readings. Descriptive Statistics N Minimum radon readings (hw7) 12 Valid N (listwise) 12 91.90 Maximum 122.30 Mean Std. Deviation 104.1333 9.39742 B. What are the data requirements for conducting a t-test for the population mean (sigma unknown)? The requirements include a random sample and either the numbers are normally distributed or the sample size is >=30 with #s that aren’t highly skewed. C. Using the estimated sample mean and standard deviation, conduct a one-sample t-test to assess whether the manufacturer’s detectors were accurately measuring the true radon count? Show the hypothesis, sample t statistic, critical t value and decision rule. Ho: population mean = 105 Sample t =- .32 Ha: population mean not equal 105 Critical t (2 tailed at 5%) = 2.20 Since the |Sample t| is < |Critical t|, we can’t reject the null hypothesis. There’s insufficient evidence to say that the population mean is not 105 (hence they are correctly calibrated). D. What does the estimated p value (also called the sig. level) of the test show? The p-value of .75 is the probability of rejecting the null hypothesis when it’s true. It shows the probability of observing samples whose means differ at least as much as this one does from the null value, when the null is true. E. Use SPSS to calculate the appropriate sample t statistic by clicking on Analyze * Compare Means * One Sample T-test, and then moving the variable name into the Test Variable box, typing in the appropriate Test Value and clicking on OK. Does the SPSS sample t agree with the one that you calculated? SPSS generates the same sample t (of- .32). 4. Assume that the CDC is concerned about exposure to lead at a particular company and randomly samples 30 employees blood. Also assume that the average blood lead level (BLL) for those sampled is 8 ug/dl, with a standard deviation of 2. Also assume that "normal" BLLs average 6 ug/dl. A. What are the data requirements for conducting a t-test for the population mean (sigma unknown)? The requirements include a random sample and either the numbers are normally distributed or the sample size is >=30 with #s that aren’t highly skewed. B. Using the estimated sample mean and standard deviation, conduct a one-sample t-test to assess whether the population mean BLL level for all company employees is greater than the "normal" level of 6. Show the hypothesis, sample t statistic, critical t value and decision rule. Ho: population mean <= 6 Sample t =5.48 Ha: population mean > 6 Critical t (1 tailed at 5%) = 1.70 Since the |Sample t| is > |Critical t|, we reject the null hypothesis. There’s sufficient evidence that the population mean BLL for this company’s employees is greater than 6 (and the sample mean agrees with the alternative hypothesis, which is needed in a one-tailed test in order to reject Ho). We’re 95% sure with a 5% chance of erroneously rejecting the null. C. What does the estimated p value (also called the sig. level) of the test show? The p-value of .000003 is the probability of rejecting the null hypothesis when it’s true. It shows the probability of observing samples whose means differ at least as much as this one does from the null value, when the null is true. Samples like this are extremely rare if the null was true. #1 Z Test of Hypothesis for the Population Proportion #3 t Test of Hypothesis for the Population Mean (sigma unknown) Data Null Hypothesis p= Level of Significance Number of Successes Sample Size Data Null Hypothesis = Level of Significance Sample Size Sample Mean Sample Standard Deviation 0.18 0.05 23 235 Intermediate Calculations Sample Proportion 0.09787234 Standard Error 0.025061626 Sample Z Test Statistic 3.277028354 One-Tail Test Absolute Critical Z Value p - value 1.644853476 0.000524589 Two-Tail Test Absolute Critical Z Value p - value 1.959962787 0.001049178 105 0.05 12 104.1333 9.39742 Intermediate Calculations Standard Error of the Mean 2.712801483 Degrees of Freedom 11 Sample t Test Statistic -0.319485228 One-Tail Test Absolute Critical T Value p - value 1.795883691 0.377672052 Two-Tailed Test Absolute Critical T Value p - value 2.200986273 0.755344105 #3 One-Sample Test Test Value = 105 95% Confidence Interval of the Difference t radon readings (hw7) df -.319 Sig. (2-tailed) 11 .755 Mean Difference -.86667 Lower -6.8375 Upper 5.1042 #2 Z Test of Hypothesis for the Population Proportion #4 t Test of Hypothesis for the Population Mean (sigma unknown) Data Null Hypothesis p= Level of Significance Number of Successes Sample Size Data Null Hypothesis = Level of Significance Sample Size Sample Mean Sample Standard Deviation 6 0.05 30 8 2 0.1 0.05 6 40 Intermediate Calculations Sample Proportion Standard Error Sample Z Test Statistic 0.15 0.047434165 1.054092553 Intermediate Calculations Standard Error of the Mean Degrees of Freedom Sample t Test Statistic 0.365148372 29 5.477225575 One-Tail Test Absolute Critical Z Value p - value 1.644853476 0.145920298 One-Tail Test Absolute Critical T Value p - value 1.699127097 3.36957E-06 Two-Tail Test Absolute Critical Z Value p - value 1.959962787 0.291840597 Two-Tailed Test Absolute Critical T Value p - value 2.045230758 6.73915E-06