Research Design, Biostatistics, and Literature Evaluation

Descriptive statistics

Measures of central tendency

Mean: Used for continuous and normally distributed data (parametric), sensitive to outliers.

Typically presented in conjunction with standard deviation.

Median: 50th percentile, used for ordinal data or continuous and non-normally distributed data

(nonparametric). Typically presented in conjunction with interquartile range.

Standard deviation

Used for continuous, parametric data

ii.

Describes variability around the mean

iii.

Approximately 95% of the data fall within two standard deviations of the mean

Range: Describes spread of data, minimum to maximum values

Percentiles: The value where a percentage of values falls below. Example: The 75th percentile is

where 75% of values are smaller.

Interquartile range: Describes the values between the 25th and 75th percentiles

Used for continuous, nonparametric data

Paired data

There are circumstances in which matching will occur naturally between data points

Duplicate measurements on one sample

ii.

Sequential measurements, such as a pre-post test

iii.

Cross-over trials

iv.

Matched samples

Two independent groups: Student t-test

Two independent groups: Wilcoxon rank sum or Mann-Whitney U

Nominal Data

Two or more categories (e.g., sex, race, treatment groups, CAM-ICU [Confusion Assessment Method for

the ICU])

Fisher exact test: More accurate for smaller data sets (fewer than five observations per group)

Groups represented by scale (e.g., Richmond Agitation-Sedation Scale score, Sequential Organ Failure

Assessment score)

Statistical tests: Wilcoxon rank sum, Mann-Whitney U, Kruskal-Wallis

A method to describe variability around a point estimate within the study population

Magnitude of difference between groups

Range of values (possible spread of point estimates)