lecture 3 Flashcards
define class
one of the categories into qualitative data can be classified
define frequency
number of obs (record feature, catgorical, like colour) into a particular class/category
define relative frequency
class freq / total number of obs in data set
define percentage
Class relative freq x 100
describe graphical methods
form tables or graphs
many types of graphical displays
summarize data = display so can be understood
describe pie chart - general
classes represented by slice of pie = circle
size of each size proportional to relative freq of diff classes in data set
describe pie chart - aspects
can put counts in the slices
better to put percent - relative freq bc more informative
can separate by region if wanna see how data is
describe bar plot - general
classes represented by bars
height of each bar either class freq or class relative freq or class percentage
width of bars not important
describe bar plot - specific
do not learn much but just way to represent
can have counts or percents, can separate by region
Describe numerical methods
class freq, relative freq, percentage
usually displayed in table
describe kidney stone treatment - general
designed experiment
Experimental units = patients with kidney stones
2 treatment groups = a and b
each patient assigned to one treatment group
describe kidney stone treatment - variables
3 =
Treatment group= a or b
outcome = success/failure of removing stones
size of stone = large/small
ALL QUALITATIVE- no numerical scale attached
describe kidney stone treatment - table
shows # of patients treated per treatment
number = frequ of success per treatment
success rate percentage per treatment
do not differ significantly but b is a bit better, seems b is better since higher percentage of success
describe kidney stone treatment - table of size
Treatment a has higher percentage of success with both small and large sized stones
success rates diff for treatment group and for kidney stone size
allocation of sizes of stones to diff groups is uneven
but overall = docs recommend b
pooling into single table disregards effect of conditional probability
describe kidney stone treatment - issues
number of patients when considering stone size are very diff
doctors tend to give treatment a to large stone and b to small stone
patients with large stones using treatment a do worse than those with small stones, even if small stones used b
but data doesnt indicate a is better overall
describe kidney stone treatment - interpretation
Relationship between treatment and success/failure changed when stone size considered
must be careful when interpreting results from study
paradox = mistaken interpretation of what number means
what does kidney stone data illustrate
simpsons paradox
Relationship between treatment method and success rate confounded by stone size
*not consequence of treatment - no influence on it
describe simpsons paradox
third variable changes interpretation of relationship between other variables
third relationship = confounding factor
Changed interpretation
can happen with 4th variable = more complicated
describe simpsons paradox uc berkeley college - variables
sex = male/female
majors
Admission’s status = admitted/rejected
each = QUALITATIVE DATA
describe simpsons paradox uc berkeley college - table of frequencies admitted
males = way higher than females
looks like bias against females
but not enough info
need number who applied = successful applicant/total applicants
describe simpsons paradox uc berkeley college - table including % admitted
males = 44.5%
females = 30.4%
seems to indicate bias against females
describe simpsons paradox uc berkeley college - table including major
when including % admitted per major = for males and females numbers pretty similar =
cannot make conclusion about bias against females based on the table results
describe simpsons paradox uc berkeley college conclusions
major = confounding factor
no department was significantly biased against females
relationship between sex and admission status confounded by major
be careful when interpreting analysis
have to account properly to draw conclusions
describe underreporting
make things appear different than they are
