research skills 9 statistical significance

Question 1

what are error bars ?

Answer 1

• Often it is the mean value that is plotted on the final graph of your results
• Error bars represent the uncertainty or variation in your data, they indicate how spread out the data is around the mean.
• They often represent the standard deviation, the standard error, the range or the confidence interval.

Question 2

what can error bars be used for ?

Answer 2

• Descriptive statistics: range, standard deviation
• Inferential statistics: standard error, confidence intervals

Question 3

when should you use error bars ?

Answer 3

• Only use error bars to display the variance in the data from at least 3 independent experimental replicates.
• Make sure you use the right error bars for the type of data
• Always give a n value to indicate how many replicates of data are presented.
• Exception – if you are testing a new technique or new protocol and you need to see how reproducible the results are

Question 4

how do you calculate fold/simple change ?

Answer 4

change = (mean experimental) - ( mean control )

Question 5

how do you calculate % increase ?

Answer 5

( final value - initial value) / initial value x 100

Question 6

what is effect size ?

Answer 6

• The effect size is a simple way to quantify the difference between 2 groups
• The effect size is used when you want to know more than just “is there a statistical difference”, when you need to know how much is the difference – what is the magnitude?
• The effect size emphasises the size of the difference

Question 7

how do you calculate effect size ?

Answer 7

( mean experimental) - ( mean control) / SD ( average from 2 groups)

Question 8

what does the effect size mean ?

Answer 8

• The effect size number calculates how many standard deviations
• This can be positive and negative

Effect size – what do they mean?

0 – 0.2 = no effect or small effect
0.2 - 0.5 = medium effect
> 0.5 = large effect

Question 9

why should standard error bars be treated with caution ?

Answer 9

• With low numbers of samples (<20) they underestimate the uncertainty
• All statistics give only a rough indication of confidence
• 1 x SEM shows that only 68% of all sample means will be within one standard error of the population mean (and 95% within two standard errors).

Question 10

what are the guidelines for a small sample size ?

Answer 10

Question 11

what are the guidelines for a large sample size ?

Answer 11

Question 12

what are the advantages of confidence intervals ?

Answer 12

• Combine numerical information on effect size, statistical confidence, and possible variation in the “real” effect size
• Ideal for simple comparisons such as treated vs control
• Now the preferred approach in clinical research and epidemiology