research Flashcards
Analyzing data - 3 ways
Relative Ratio - RR
Odd Ratio - OD
Risk Difference ARD - absolute risk
do you need a confidence interval?
Yes - you need to be certain within a certain degree - 95% is standard (actually 95.5)
all measure of effect are uncertain
risk ratio - table for four
people who have dogs smile more -
35/50, 12/28
.7/.525 = 1.33
intervention goes first
if you are comparing two active interventions - you can choose which goes first
what does 1.33 mean?
> 1 increases the risk of the event occurring
if =1 - there is no affect
in RR - where number is >1 - means the intervention (have a dog) increases of risk of smiling
if <1 0 means having a dog dog not increase the risk of smiling
Odds of an Event?
OR >1 - intervention increases the odds
if OR = 1 - no difference
figure out the odds of both groups, divide the two and get the odds ratio
35/15; 12/28
- 33/ .428
- 44
odds are that if you have a dog, you are five times as likely to smile
What does 5.44 mean?
Can be expressed in a couple of ways
odds are 5 times more that if you have a dog you will smile
and
can be expressed as a percentage - 544% higher than the odds in the control group
Risk difference - an absolute measure - not a ratio -
RD>1 intervention increases the risk of event
RD<1:
intervention reduced the risk of event
?????
HERE _ point of no effect is ZERO
here - could I say?
having a dog increases the risk of smiles
35/50
12 /40
.7 - .525
subtracting
= .175
??
two ways to express -
- percentage points
percentage change
percentage point change
- natural frequencies
negative skew vs positive -
negative skew - tail is on left
NEGATIVE IS little and MEAN
mean < median < mode
MEDIAN is always in between these two with either skew
positive skew?
Big meanie is RIGHT
tail is on RIGHT
mean > median > mode. Generally speaking, extreme values in skewed distributions affect the mean,
