exam 3- ATTR 425

nominal

weakest and most basic level, provides a name, no order or rank, no quantitative data, frequency only

ex: polls- relating to sex, ethnicity, state, political party, employment status

• typically demonstrated in visual graphs such as bar graphs, pie charts, etc,

common mistake with nominal data

running mean, median, or mode- frequency data. you shouldnt be able to compute this

ordinal

uses labels, has order, value between labels is meaningless

frequency, mode, median, range

ex: pain scale, borg scale, likert scale, age range groups,education level

• we do not know the value or distance between the order, only know the order.

• with age groups- age range groups are ordinal, age is alone- this will be ratio level data

interval level

higher level of measurement, provides labels, has order, values between labels is consistent, no absolute zero, can calculate MCT, range, stdev, variance

• no absolute zero- zero should mean the absence of that thing.

ex: temperature, range of motion, time (within 12 hr clock)

ratio level

highest level of data, provides all info of nominal , ordinal, and interval

has a zero that can be meaningful

ex: weight, height, length of time, length

parametric data

fall within a normal distribution, allow researchers to draw conclusions mathematically; a normal distribution is symmetric around its mean

• think parameter- allows us to draw conclusions or parameters about a group of things

• INTERVAL AND RATIO

non parametric data

nominal and ordinal data

measures of central tendency

revolve around mean, median, and mode- what they are

these values cluster around the center of normal distribution

arthimetic mean

average of all scores, can be skewed by extreme scores- or outliers

median score

falls in middle of a distribution, not as easily skewed by extreme scores

can be a better data indicator if u have large groups with outliers

mode

most commonly occuring value in a distribution of numeric data

goal of quantitative research designs

to establish relationships between measures of interest

Quasi- Experimental

• attempts to establish a cause and effect between variables

• lacks random assignment

• often very clinical research

• the IV is not manipulated to change the DV

• fails to control for all variables

Sampling: what is generalizability

It asks whether the reported measure or outcome can be reproduced

the reader needs to know if the results reported are reproducible with particular patient

it is influenced by sample size and population representation

Internal validity

design of study is measuring what it is intended to measure

interpreting the data for cause and effect

External Validity

ability to generalize results to sample population

generazibility of findings and conclusions

parameter

characteristic of an entire population

ex: VO2, FIIT principles, etc.

1- test samples using statistics

statisitc

characteristic of sample - used to establish a parameter

a priori

from the earlier

a posterior

the later

descriptive statisitcs

condense large amts of data into easy to interpret fashions

T- test

compares two groups with a single measure

ex: before and after treatment

limited to one dimensional comparisons: usually do not want to run multple T- tests on the same data

ANOVA

allows to compare more than one set of means at a time

decreases the fear of committing a family wise error

Independent Samples t-test

used when independent variable has 2 levels

one way ANOVA

used when independent variable has three or more levels

regression

used to predict the balue of a DV based on known relations between independent variables

null hypothesis

result of your study that will have no effect on what you are studying

Treatment A= Treatment B (no effect)

• treatments have the same idea, usually in research you do not want this to happen

Type I error

Null hypothesis is true and the researcher rejects the null hypothesis

False positive - more common in research

When you say there is a difference but there is not

You usually do not want to start a research study saying there is no difference- which is why it happens more commonly

Type II error

Researcher fails to reject the null and the null hypothesis is false

False negative

statistical power

probability that we will reject the null hypothesis when that hypothesis is true

probability you will get it right when that difference is true

ways to influence power

increase sample size, increase effect size, less variance in subjects, raise P-value

P- value

The likelihood you will commit a type I error at the end of your study - when you say there is a difference when there is actually no difference

typically set .05 but can be set at any level

does not denote effect size - it only gives the indication that difference between groups is likely not due to chance/ error

95% confidence interval

95 % of data falls between upper and lower bounds of a bell curve