ask a question/ identify problem

hypothesis-proposed answer to question

literature review

plan

research

a proposed answer to question

should be stated in positive terms

ex: 2nd year DH students noted heavy plaque on the posterior teeth of 1st graders but less plaque on anterior of 1st graders.

hypothesis for this: six and seven year children brush anterior teeth for a longer period of time than posterior.

a statement about between 2 groups where there is no difference.

example:six and seven year old children showed no difference in the effectiveness of plaque removal with brand x toothbrush

 who will participate?

who will gather data?

how will the data be gathered?

when will they gather data?

CALIBRATION MUST HAPPEN

systematic reviews  (secondary research)

randomized controlled studies

cohort studies

case series

case reports

ideas, editorial, opinions

animal research

in vitro (test tube) research

trial run

prepares for major study

method of sampling that provides the most external validity or degree to which the results of study can be generalized to settings

-each member fo population has equal chance to be in study.

-prevents possibility of selection bias by researcher

subdivision of a population with similar characteristics (called strata)

random selections of two or more strata is another way of defining stratified sampling

selection using every (nth, or so many patients)

(example every 3rd person in the phone book)

could have bias

sample based upon personal judgement of those who would most represent population

group of individuals who are most readily available to be subjects

could have bias

group in study who does not recieve treatment

placebo effect

provides baseline against which the effects of intervention on the experimental group can be measured.

what you give to the experimental group.

is controlled by experimenter

what is new or different

it is what happens or not

depends on or is caused by independent variable

*it is the outcome variable of interest

*it is the effect of the independent variable (it happens second)

example of variables:

the independent is in blue

the dependent is in red

 has one of a limited set of values and are counted only in whole numbers

example: hair color, gender, political preference

measurments made from a particular value within a defined range

(example: temperature, scores on a test)

things that change

Quantitative in nature

measurement of names categories with no order

scores that cannot be heirarchally ordered

example: females placed in category a. males in category b. (one sex is not better or less than the other)

nominal scales are put into groups or classes

ranked order is used

example:

-stage I cancer

-stage II cancer

-stage III cancer

Stage IV cancer

other examples mohs hardness scale

have a common unit of measurement between each score, but no true zero

(example is temperature)

quantitative: mean, median, mode, standard deviation

have equal intervals between the variables but there is a meaningful zero point.

example: height / weight

mass

length

time

plane angle

energy

electric charge

nominal scale

ordinal scale

frequency-tabulating a group of scores

central tendency- scores tend to clump together

shape-whether scores pile up together (skewed vs. normal)

variability-how much scores are spread out

used to describe and summarize data

objective: communication results without generalizing.

may use central tendency or measures of dispersement

 list all scores

 put in ascending or decending order

slashes used

example

score                           frequency

10           /         1

this means 1 person got this question right out of 10 people

grouped frequency

distribution

uses range

example: 10 intervals

if range is 96, highest, 27 lowest, then the range is 69

interval width= range/ # of intervals

scores are clumped together (bell curve)

 3 methods of computing

-mode

-median

-mean

most frequently occuring score

least reported measure of central tendency/ least stable

greatly influenced by a few scores

can be unimodal, bimodal, or multimodal

the point that divides the distribution in half

the midpoint between all grades

if odd # of scores there is a difinitive median

if even # add two middle scores and divide by 2

whn distributions are skewed, the median represents central tendency.

average of all scores

take all scores/ how many scores are used

x=sum

N

affected by extreme scores

can be pulled in either direction by very high or low very scores

 whether or not scores pile up at one end (normal vs skewed)

in normal mean, median , mode are equal

in skewed mean is pulled toward the tail, and median falls between mean and mode.

mean is highest

median has middle value

mode is lowest (most frequent score)

more students had low scores indicating test too difficult

mode is highest

median is middle

mean is lowest

most students do well, but a few scores bring average down.

how spread out are scores

 use range= highest - lowest

example:

high score 100

low score 50

range = 50

take square root of variance

SD is influenced by all values

large SD= large variability, or heterogeneity

small SD= small variability or homogeneity (scores cluster close to mean)

this measure is most appropriate for normal distributions

measure of relation between 2 or more variables

look for clumping of dots on xy axis.  if lots of clumping, they are closely related.

mean and SD has been determined, correlations are used to see if one variable my be relatied to a variation in another variable

2 variables increase

example of positive

students grades go up when they study alot

x axis up

y axis up

no parameters, parameter free

(have no idea how data will turn out)

aka students test

compares mean of two groups (group A vs group B)

does experimental group have a higher mean?

analysis of variance

allows comparison for 2  or more groups

compares variability within groups with variability between groups as a whole.

nonparametric inferential(draw a conclusionand make prediction) stat technique

most common test for significance of the relationship between categorical variables

degrees of freedom (2 choices)

researchers can calculate the probability of statistics being correct.

researchers usually us a 95%-99% confidence interval.  this indicates the results will be wrong 5 times in 100.

p value states how likely it is that a study could have come to a false scientific conclusion.

calculated according to sample size, difference of means in control and experimental groups, and standard deviation of distribution.

acceptable p values will be less than .05% (the smaller number the more significant the study)