EXSC 320 - Exam 2

descriptive statistics

measurements obtained from a sample that seek to describe the data set (ex. height, weight)

Frequency distribution

how often a number appears, normal distribution, positive skewness (left), negative skewness (right)

Kurtosis

related to variability or peakedness

mesokurtic

normal, bell shaped

leptokurtic

homogeneous, very narrow and peaked

platykurtic

heterogenous, much variability

measures of central tendency

mode, median, mean

mode

most frequent score, terminal

median

50 percentile of a group

mean

sum of x/n, gives weight to each score (x bar = sum of x/n)

measures of variability

range, variance, standard deviation

range

high and low score

variance

measures the distance each score is from the mean

s² = sum of (x - xbar)²/n-1

standard deviation

s = sqrt(variance)

standard scores

allows for meaningful comparison to be made between differing data sets

z scores

standardized scores expressed in relation to standard deviation units on a normal curve

z = (x-xbar)/s

t scores

10((x-xbar)/s) + 50

correlation

represent a relationship between 2 variables (-1 to +1)

perfect correlation

1.0

very high correlation

.80-.99

high correlation

.70-.79

moderate correlation

.40-.69

fairly low correlation

.30-.39

low correlation

.20-.29

very low correlation

<.19

no correlation

0.0

2 reasons correlations can be negative

opposite score scale

true negative relationship

opposite score scale

an increase in 1 number corresponds with a desirable decrease in the other (ex. lower 5k and higher VO2 max)

relationship

correlation

prediction

regression

simple regression

y is predicted from x (y’ = a + bx)

multiple regression

y is predicted from more than 1 variable

t-test

trying to determine difference in means

significance level (alpha)

determines if there is a statistically significant difference in the mean (.05)

dependent t-test

compare means for 2 variables in a within subject design

(1 group, 2 variables)

independent t-test

2 groups and 1 variable

one sided/tailed

predicting a direction (one is greater than another), cut significance in half

two sided/tailed

predicting there is a difference (not greater than another)

type 1 error

no relationship exists but you conclude there is a relationship

type 2 error

a relationship exists but you concluded there is none

ANOVA

analysis of variance

repeated measures ANOVA

within subject design with design of greater than 2 variables

(1 group, > 2 variables)

one way/univariate ANOVA

between groups

(> 2 groups, 1 variable)

population

an entire group from which the sample is selected

sample

the subgroup of a population from which data is collected

sampling

a process by which you choose the sample

sampling frame

accessible population from which the subjects are selected

measurement validity

how well a tool measures

relevance

is it representative of the topic (face, content)

empirical

validity based off collecting and analyzing data (criterion, research related)

cirterion

comparing measurement tools against a criterion

research related

using research to determine validity, all research not criterion related (construct, convergent, discriminant)

construct

take 2 groups on opposite end of the scale and determine difference in scores

convergent & discriminant

use correlations - high positive correlation is convergent, no correlation is discriminant

face

on the surface does the tool appear to measure what it says that it does

content

examine the content to see if the attribute of interest is adequetly covered

measurement reliability

consistency of measurement (good or bad)

high validity, high reliability = desired

sampling error

the extent from which sample values could deviate from entire population

probability sampling

a sampling technique based upon random processes in which the probability of selection is known (simple, stratified, cluster, systematic)

simple random sampling

a type of probability sampling in which every element in the population has an equal chance of being selected

stratified random sampling

a population is stratified into particular groups and subjects are randomly selected from each group

cluster sampling

probability sampling in which subjects are chosen from naturally formed clusters

systematic sampling

a type of probability sampling where subjects are chosen at specific intervals

non probability sampling

subjects not randomly chosen

factors impacting sample size

1. sampling error is inversely related to sample size

2. descriptive and correlational studies require a larger sample size

3. heterogeneous populations require a larger sample size in relation to homogeneous

4. when samples are divided into smaller groups, the initial sample should be large enough that subgroups are of adequate size

5. practical factors such as subject availability and cost are legit considerations when determining sample size

4 levels of measurement

nominal, ordinal, interval, ratio

nominal

numbers indicate categories, do not reflect magnitude

ordinal

numbers indicate rank order, do not indicate magnitude

interval

numbers represent equal units between measures and magnitude can be determined, no true zero

ratio

numbers represent equal units between measures and magnitude can be determined, is a true zero