Biostatistics Unit 1

Statistical inference

creating ranges from samples in the hopes that those ranges contain the true number

Two types of variables

numerical and categorical

Types of numerical values

discrete (can take only integer value) or continous (can take any value within a range)

Types of categorical values

nominal (name the categories) and ordinal (have an intrinsic order)

Graphing categorical variables

pie charts, and bar charts/graph

Graphing numerical variables

histograms and box-plots

What information do you get from a histogram?

An idea of the center of the data, the spread of the data, the shape of the data and the modality.

Symmetric shape

bell-shaped, uniform, neither

What direction is skewed data?

To the opposite of where the data is heavy. Look for long tail

Modality

how many peaks are in the data

What information do you get from a box-plot?

The exact center of the data, how the data is spread, the shape of the data (are whiskers long or symmetrical?), outliers. Cannot see modality

What is the five number summary of data?

Minimum, maximum, median, and q1 and q3

bivariable

looking at two variables at the same time

How do you graphically describe bivariable categorical data?

Mosaic plots, stacked bar charts

How do you graphically describe bivariable numerical data?

scatterplots

How do you numerically describe bivariable numerical data?

correlation coefficient. If cor = 0, no linear correlation

What is the error in a scatterplot?

The difference between the LOB and the points

How do you numerically describe categorical data?

frequency tables

How do you numerically describe numerical data?

Mean, Standard deviation, median, and inner quartile range

70-95-99 rule

99% of data should be within 3 standard deviations from the center

Statistics

science of collecting, and organizing data from a sample with a focus on making conclusions about populations

When can we generalize results from a sample to a population?

When the sample is randomized (random from the entire population) and/or representative

Experimental sampling

data needs to be divided in at least two groups (control and treatment), Cannot make generalizations about a population

Observational sampling

data is just collected with no manipulation, Cannot be used to draw conclusions for cause and effect

Convenience sampling

Data is gathered however possible, like polling people on the side of the road

Simple random sampling

having a list of the entire population and sampling randomly. Sample can end up very unrepresentational

Stratified sampling

splitting the population using one criteria. For example, in a population of 100, 30% male and 70% female, selecting 3 random males and 7 random females to make it random and representational

Clustered sampling

Stratification done multiple times and get many stratified samples (clusters). Picks a couple of clusters to study and combines

Multistage sampling

creating a cluster sample and then sampling from it as if it is a population to form the final sample