Unformatted text preview: Biostatistics Objectives Define biostatistics Differentiate between descriptive and inferential statistics Describe qualitative and quantitative data Describe nominal, ordinal, interval and ratio data Points to consider when selecting a statistical test Biostatistics Procedures for condensing, describing, analyzing and interpreting large sets of health relevant information Helps us make sense of information that may seem random Select the statistic according to the information generated and the questions to be answered Selecting a Statistical Test: Selecting Points to Consider Points Target population & variables to study How sample is selected from the target population The generalizations you want to make about the target population Level or type of data to be analyzed How data are distributed Populations and Samples Sample is a subset of a population targeted for an epidemiological investigation Selecting a sample: Probability or random Nonprobability Descriptive vs Inferential Statistics Descriptive: Indicators used to represent quantitative summaries of numerical information Rates Frequencies Ratios Descriptive Statistics Methods used to organize and produce quantitative summaries of numerical information, primarily by means of: Tables Charts Graphs or Diagrams Descriptive Statistics In epidemiology, the most commonly used descriptive statistic is a RATE Other descriptive statistics include: Frequencies Indices Ratios Proportions Measures of central tendency & dispersion Descriptive vs Inferential Inferential: Used to make generalizations about a larger population on the basis of information derived from a representative subset of the same group Differences between 2 magnitudes Relationship between 2 variables Appraisal of 2 or more interventions Trend of a variable of interest Inferential Statistics Commonly Inferential used in Epidemiology used Nonparametric Tests Variables are NOT normally distributed “Distribution Free” ChiSquare, Wilcoxon, & Mann Whitney U Tests
Nominal (categorical) Ordinal (ranked) Inferential Statistics Commonly Inferential used in Epidemiology used Parametric Tests Assume the variable is normally distributed z, t, & F tests The data are quantitative: Interval or Ratio Data Information (reports of observations) collected on variables identified with the objects of interest in a sample or population Two Types of numerical data basic to biostatistics Qualitative Quantitative Categorical “All or none” information measured with nominal or ordinal scales Nominal Information that can be classified into mutually exclusive categories Dichotomous: if there are only 2 mutually exclusive categories Typical analysis: Logistic Regression Categorical/Quantitative Ordinal: Information on variables that may be counted, but these counts may also represent gradations of the variables Usually rankordered, but may not be able to tell how much difference there is between categories Examples: Likerttype Scales, ranging from strongly agree to strongly disagree; Asking a person to describe their health status on a scale from 15; Ordinal level of measurement Often used & misused in social science & medicine Data are rankordered, yet analyses used require interval level of measurement Result: True relationship can be underestimated; Interpretation of results difficult Common Analyses: Logistic regression; Chi square; MannWhitney tests “True” Quantitative Derived from a count or a standard measurement and has a frequency distribution Measured with interval or ratio data Examples: height & weight; other physiological variables Quantitative Interval Measured in standard units Any given differences between two numerical values has the same meaning Most standard statistical analyses require this level of measurement to be accurate; ex: linear regression, ANOVA Quantitative Ratio Measured in standard units Scale they are measured on has a true zero point that represents total absence of the variable E.g. volume, mass Rarely applicable in social science and medicine Quantitative Discrete vs Continuous Discrete: limited number of values (e.g., number of children in a family) Includes nominal & ordinal measures Continuous: infinite number of values (e.g., weight) Includes interval & ratio measures Qualitative Text analysis Interview & focus group transcripts Goal is to identify themes within text related to the interview questions Provides contextual information often missing from quantitative data Aids in interpretation of quantitative data Importance of Understanding Importance Types of Data Types Important when you are selecting a statistical test e.g., age vs gender e.g., selfperceived health status Overview of Statistical Tests The relationship between: Level of measurement Choosing a statistical test (nonparametric or parametric?) Hypothesis testing and pvalues: Choosing a Type I error rate convention:
p<.05, or p<.01 How does sample size affect pvalues? Confidence Intervals: much like the margin of error concept used in polling Statistical Significance Test the research question State the Null Hypothesis, Explicitly or Implicitly Null hypothesis: Differences are due to sampling variability only Alternate hypothesis: Differences due to sampling variability AND true differences in population parameters Probability & Hypothesis Testing Choose a significance test (pvalues) P-Values What is the relationship between: Sample size? Confidence Interval? Pvalues? Sample size is part of the formula for calculating statistical significance. Large sample sizes are more likely to generate statistically significant, but not practically significant results ...
View Full Document
This note was uploaded on 04/12/2009 for the course HED 10520 taught by Professor Edmundson during the Spring '08 term at University of Texas.
- Spring '08