Review Question - QID 103891

When analyzing continuous data from two separate, normally distributed samples, the most appropriate statistical test is an independent two-sample t-test.

When analyzing data from clinical studies, the characteristics of the data generally determine which statistical tests can or should be used. In this study, the systolic blood pressure (SBP) has been measured on 1000 patients, 500 who have a low salt diet and 500 who have a regular diet. The mean and standard deviation are presented in Figure A. Despite the fact that the low salt dieters have a lower SBP than controls, this data in and of itself does not answer the question of the study, which is: do patients on low salt diet have different SBP than controls? That is, the fact that these 1000 patients happened to have produced this data, does not mean that another 1000 patients similarly sampled will also yield the same result. To answer the study's question, one must analyze the data with a statistical test. As SBP is a continuous variable and two groups have been sampled, the most appropriate test is a two-sample t-test.

Rosner delineates the situations in which the two-sample t-test can be used: 1) The data come from two samples of different groups, 2) The samples are independent of each other, 3) The underlying distribution of each group is normal or the central-limit theorem can be expected to hold, and 4) The parameter being compared is the mean of the groups. One can perform statistical tests to determine whether the normality assumption is met, or one can follow the general rule of thumb that samples of at least 30 observations are sufficiently large enough for the central limit theorem to hold.

Figure A presents the summary statistics of the study. Means and standard deviations of the SBP for each diet group are provided.

Incorrect Answers:
Answer 1: An ANOVA test is generally used to compare the means of more than two independent groups. An ANOVA comparing only two samples will yield an identical p value as the t-test, but it is more difficult to calculate and thus the less ideal choice.
Answer 2: One-sample t-tests are used to compare the mean of a single group to a given value. The current analysis has two groups and thus a one-sample test cannot be used.
Answer 4: Paired t-tests are used to compare means of samples which are inherently correlated, e.g. change in blood pressure within individuals on a given diet. The two separate arms of the current study prevent this analysis.
Answer 5: Chi square tests are most frequently used for non-continuous data. As blood pressure is a continuous variable, this analysis should not be used.