Statistics-Biomedical Sciences.txt

A field of study that involves methods for describing and analyzing data. It reduces uncertainty and provides for better decision making.

Universe of cases or subjects of interest to the analyst. People, thins or concepts

An observable subset of the population. It needs to mirror the population.

Random and Non Random

All units have an equal chance of being included in the sample.

• 1. Using a table of random numbers
• 2. Computer generated random samples
• 3. Software selected random sample

• -Assign all elements a number, in a class of 25 - assign 1 - 25
• -Determine the sample size example: 5
• -Use a table to assign random numbers to provide 5 random numbers between 1 and 25

• -Produce names of population
• -Determine sample size (5)
• -Divide the total (25) by the sample size (5) = 5
• -Take every 5th name on the list for inclusion in the sample
• -If the 5th person refuses, the analyst must begin the count to 5 again.

Divide the number of STRATA (groups) that share similar characteristics. Draw random samples from each Stratum.

Sampling based on selecting clusters from a population and then sampling from those clusters. Examples: geography - rural, suburbs, city

Cluster samples only include a subset of the clusters. Stratified samples include all of the strata. Stratified samples allow for more precision.

• 1. Convience samples - surveying the first 10 people in a parking lot
• 2. Volunteers- American Idol
• 3. Judgemental sample- a sample based on expert judgment
• 4. Quota sample - convience sample designed to provide a certain distribution

The difference between the sample and the larger population that is due to pure random chance .

As sample size increases sampling error decreases .

The differences between a sample and the population that are not do to pure random chance.

Unlike sampling error, sampling bias will not decrease as your sampling number increases .

• -A group that is under represented in your samples.
• -A group that fails to respond to your survey non response bias
• -A group that self select so as the sample, American Idol.

Inaccuracy or miscalculation of the observation , caused by unclear questions, leading questions, questions containing social desirability componant.

Does the instrument measure what it intended to measure.

Does the instrument provide constant results over repeated measurements.

• 1. Face - does the anayst have confidence in the measuring instrument
• 2. Content- concerned with the sample population representatives-
• 3. Correlative - the results have a high correlation to other established measures of validity
• 4. Predictive- the resilts should be able to successfully predict outcomes gre = success in a graduate program

Results that can be readily generalized to the larger population are said to have external validity.

Did I measure what I claimed to measure by eliminating all confounding variables

• 1. History - external events that produce an effect that can be confused with the outcome. school program success vs economic boom happening at same time
• 2. Maturation - internal factors that can be confused with outcome - treated allergies that resolve over time, due to tx or due to growth of child
• 3. Testing- measuring a person that can produce the effect confusing the outcome- Stalins arrival improves productivity
• 4. Instrumentation - changes in the measurement tool
• 5. Statistical regression to the mean - selection of a group due to their deviance from the mean - odds are that next measurement that group will regressed to
• mean.
• 6. Selection bias
• 7. Experimental mortality- subjects dropping out of study will change the composition of the sample.
• 8. Selection-Maturation Interaction -any bias in selection will interact with maturation to produce a greater effect than maturation alone

• 1. Pre-experimental - policy is changed and later a decision is made to evaluate the policy
• 2. Quasi Experimental - uses a comparison group. Ex: impact of affirmative action on female employment in shipyards.
• 3. Experimental -includes randomization componant. Participants randomly selected and randomly assigned to experimental or control group.

Real world clinical trials often may have a drug that is valid in a controlled setting (high internal value) but not effective in the real world where patients don't follow direction (low external validity)

• 1. Nominal - catagorizing information. Hair Color: 1= blond 2= brunette 3= other
• 2. Ordinal - ranked in order of some rype of continuum. 1= strongly agree 2= agree 3= neutral 4= disagree 5= strongly diagree
• 3. Interval - regular numbers where distance between the numbers is the same and all numbers anchored by an arbitrary zero - IQ, Temp, Test scores
• 4. Ratio Scale distance between points is equal and anchored by a non arbitrary zero. Hourly wage, height, weight, age, miles driven in a day

• -Gender
• -Ethnicity
• -Marital status

• -Movie ratings
• -Scio economic status
• -Rating of meat in the store
• -Rank order of anything

• -Degrees F OR C
• -Most personality measures
• -Intelligence scores

• -Annual income in dollars
• -Distance as measured in miles, inches, centimeters etc..

Counting

Gretaer than or less than

Addition and subtraction of the scale values

Multiplication and division of scale values

• 1. Mean
• 2. Mode
• 3. Median
• 4. Trimmed Mean

A measurement of central tendoncy that is equal to the score that occurs most often in the distribution.

The score that divides the distribution in half. =(n + 1) / 2 will identify the positon of the median

The arithmetic average of scores. Sum the scores and divide by the number of scores.

Mode and Median