Wk 3- Biomed Skills- Data Interpretation and Analysis Flashcards
Types of data presentation
Tables
Graphs
Box-and whisker plots
Bar graph scatter plots
One-way scatter plots
Two-way scatter plots
Features of tables
Numbered
Descriptive title
Column titles bolded
Footnotes
Types of tables
Freq distribution
Relative freq
Benefits of tables
Present raw data and allow own interpretations
Features of graphs
No title
Figure legend
Labelled axes
Types of graphs
Bar charts
Histograms
Benefits of graphs
Visualise data
Show trends
Features of box and whisker plots
5 no. summary
- Min
- Lower quartile
- Median
- Upper quartile
- Max
Benefits of box and whisker plots
Show range and distribution
Features of bar graph scatter plots
- = Stat analysis P <0.05
** = Stat analysis P <0.01
Features of one-way scatterplot
Control on LHS
Dots represent inds
Benefits of one-way scatterplot
Show patterns better than bar graphs
Benefits of two-way scatterplots
Show correlation and distribution
Population
Complete set of events
Often not viable
Sample
Set of observations from pop
Can become pop
Sampling methods
Representative samples
- Pop
- Representative sample
- Biased sample
Convenience samples
Systematic sampling
Random samples
Stratified random sample
Self-selected samples
Systematic sampling
Inds selected at intervals
Random samples
Inds noed then selected randomly
Stratified random samples
Inds no.ed then randomly selected
Self-selected samples
Can intro bias
Effect of sampling method
Dif methods intro dif bias
Example of sampling bias
Australian Schizophrenia Research Bank 2007 studied >1000 inds
- NSW, QLD, VIC selected via advertising
- WA, QLD selected via inpatient services
What is the ideal sample size based on?
Type of study, i.e. the dif b/n control and case
Sampling method
Accuracy needed
Why do larger sample sizes improve accuracy?
Allow smaller difs b/n cases and controls to be observed
Reduce effect of bio var
Allow larger dif to be seen b/n case and control
Sampling error
Dif b/n true parameter and sample
Types of data distribution
Direct correlation
Normal distribution
Bimodal distribution
- or + skewed
Descriptive statistics
Summarise data
Central tendency
Most typical score
Measures of central tendency
Mean- Sum of scores/no. of scores
Median- Middle score
Mode- Most common score
Dispersion
Spread of data
Measures of dispersion
Range- Highest score - lowest score
SD
What should be done w/ outliers?
They are real results and should be included unless a mistake occurred
Regression line/ line of best fit
Linear
Describes relationship b/n vars
Coefficient of determination (R2)
How well the line predicts the relationship or how close dots are to the line
Measures accuracy
Out of 1
Correlation coefficient (r)
Strength and direction of data
From -1 to 1
Aim
What you’re trying to do
Hypothesis
Proposition for exp. based on evidence
Able to be rejected
2 types of hypothesis
Null hypothesis (H0)- There is no dif b/n groups. Not what funders want to hear thus not usually written
Alternate hypothesis (H)- There is a dif
Hypothesis testing
Determines probability of rejecting hypothesis
Steps in hypothesis testing
- Propose H
- Propose H0
- Try to reject H0, i.e. prove thast the samples are dif based on reasonable prob
- Try to prove H to confirm result
- Find dif b/n control and intervention groups
- If prob that the dif b/n groups could occur w/o intervention is <0.05, reject H0 because of the statistically significant effect
Things to consider when coming to a conclusion
Did the exp have enough power to detect a dif?
Was the sample large enough?
Consider effect of gene var
