Mar4613 Exam Iii

Front 1

Nominal Data

Back 1

Numbers serve ONLY as LABELS for classifying data, CATEGORICAL data

-MODE

Front 2

Ordinal Data

Back 2

-Numbers have numerical significance but REVEAL no STRENGTH of DIFFERENCE, CATEGORICAL data

-MODE MEDIAN

ex rank your fav soda

Front 3

Interval data

Back 3

-Distance between numbers represent EQUAL VALUES

-MEAN MODE RANGE MEDIAN

ex income ranges

Front 4

Ratio Data

Back 4

-BEST KIND
-Has absolute zero point
-difficult to collect
-METRIC data

-EVERYthing plus st dev equality of ratios

ex income in exact $

Front 5

Hypothesis Ho & H1

Back 5

Ho- Null Hypothesis, Statement of no effect/difference
ex.age has no effect on water intake

H1-Alternative, some difference/effect is expected

ex younger consumers consume statistically more water than older

Front 6

Type I and Type II errors

Back 6

Type I- False positive
ex.saying age has effect when it doesn't (WORST ERROR)

type II-False negative
ex.saying age doesn't make a difference when it does

Front 7

Dependent vs Independent Variables in crosstabs

Back 7

Dep-What your trying to predict (ROWS) crosstabs

Indep-What you use to predict, demographics? (COLUMNS) crosstabs

Front 8

Crosstabs overall significance

Back 8

-cross classifies one variable against another

-chi squared value gives overall significance (.02=98%)

Front 9

Degrees of freedom in Chi square

Back 9

(row-1)(column-1)=df

Front 10

Crosstabs adjusted residual value

Back 10

-Adjusted residual must be >2 or < 2 to be significant

-if not high enough, increase sample size

Front 11

chi squared x^2 value

Back 11

If X^2 is > or = to the critical value, we REJECT THE NULL, there is a RELATIONSHIP BETWEEN VARIABLE

-ORDINAL data or less

Front 12

T test basics

Back 12

-tests for a significant difference between means

-If SD(variance) goes up, significance goes down
-if sample size (N) goes down, significance goes down

-t tests account for VARIABILITY and SAMPLE SIZE

-INTERVAL

Front 13

One sample t test

Back 13

-tests mean of a single variable against a given value

-if T value is > than critical value, its SIGNIFICANT (REJECT THE NULL, introduce product)

Front 14

Independent sample t test

Back 14

-compares differences in the means of 2 independent groups(ROWS in SPSS)

-comparing average responses ACROSS PEOPLE

Front 15

Degrees of freedom calculation in T test

Back 15

N1-N2-K=df

K=2 (ALWAYS)
N1-group 1 sample size

Front 16

Paired sample T test

Back 16

-compares mean differences in 2 variables

-comparing average responses to questions from SAME RESPONDANTS

Front 17

Lavene's test in independent sample t test

-Did we violate the constant variance assumption?

Back 17

-tests whether groups have similar variance

-assumption
1.interval data
2.constant variance
3.normal distribution

Front 18

Levenes test significance?

Back 18

if p is > or = to .05 read top line

if p is < or = to .05 read bottom line (its SIGNIFICANT)

P is SIG in levenes table in SPSS

Front 19

Anova basics

Back 19

-tests for significant difference between the means of MORE THAN 2 GROUPS
(t test is only 2 groups)

-If F is > than critical value we Reject the null(there is SIGNIFICANCE)

Front 20

F test Anova

Back 20

F=(variance BETWEEN GROUPS)/(variance WITHIN GROUPS)

-as F increases we can reject the null

Front 21

Bonferroni test in Anova

Back 21

-Multiple comparisons table in SPSS, single out 2 choices and compare significane

-Must be <.05 to be considered significant (95%)

Front 22

Degrees of Freedom in Anova

Back 22

(# of groups-1)/(samp size - # of groups)

Top of table

side of Table

Front 23

Regression basics

Back 23

-Used to try and predict/explain the variance in MULTIPLE INDEPENDENT VARIABLES

-Need interval for dependent and interval/binary for independent

Front 24

3 reasons to fit a regression model

Back 24

1.predict the dependent variable for a new set of data(ex predict sales)

2.Assess the explanatory power of the independent variable(s) (ex how is price effecting sales, elasticity)

3.Identify the subset of measure that is more effective for predicting the Dependent variable

Front 25

Regression and correlation relationship

Back 25

-Correlation is how much two variables change TOGETHER

r=1 means the 2 variables are perfectly correlated (moving in same direction

Front 26

R^2 in regression

Back 26

-gives the explained variance of the dependent variable by the independent variable

-AKA coefficient of determination

ex R^2=.80 we can explain 80% of variation in food sales (DV) is explained by coupons (IV) (not adjusted)

Front 27

Simple regression

Back 27

Involves ONLY 1 Independent variable

Front 28

Multiple Regression

Back 28

-involves 2 OR MORE independent variables

Front 29

Stepwise Regression

Back 29

Enters IV into the model one at a time but also tests IV's at each step for possible removal

-Fixes Multicollinearity

Front 30

Multicollinearity in regression

Back 30

-problem when IVs are highly correlated AMONG THEMSELVES

too much of the explanatory power of 2 IV's may be shared

VIF value of 4 or more is to high=problematic multicollinearity

Front 31

4 assumptions of regression to not violate

Back 31

1. Linearity, NOT linear curve (rainbow)

2.Independent Errors, Not straight line, same direction (arrow)

3.Constant variance, narrow to wide (shotgun blast)

4.Normality of error, line should fit on line

Front 32

Key driver in regression

Back 32

-B (beta weights) highest is most important

Front 33

Regression equation

Back 33

-Coefficient table in SPSS, bottom right box

y=Bo+B1(x1)+B2(x2)+error

Front 34

Multi dimensional scaling

Back 34

AKA perceptual mapping

-non metric (RANKINGS) or metric (PREFERENCE RATINGS)

Front 35

Disparitys in multi dimensional scaling

Back 35

disparity-differences in computer generated distances (reproduction) and the actual input provided (reality)

Front 36

R squared in multi dimensional scaling

Back 36

R squared-proportion of variance accounted for by the MDS procedure
>.6 is considered acceptable

degenerate solution-in

Front 37

stress in multi dimensional scaling

Back 37

stress-proportions of disparites to fit
0-1 where 1 equals a perfect fit

Front 38

Degenerate solution in multi dimensional scaling

Back 38

invalid solution due to:

1.inconsistencies in data

2.too few compared objects compared to dimensions
-2 dimensions need 9 object comparisons

-looks like a circle or one big cluster

Front 39

Cluster analysis

Back 39

used to identity natural segments (“clusters”) in a sample/population

-clusters that are homogeneous within and heterogeneous across.

-metric or nonmetric

Front 40

agglomerative vs divisive cluster analysis

Back 40

Agglomerativebegin with each element as a separate cluster and merge them into successively larger clusters (BOTTOM UP)

divisivebegin with the whole set and proceed to divide it into successively smaller clusters (TOP DOWN)

Front 41

Euclidian distance measure in cluster analysis

Back 41

"ordinary" distance between two points that one would measure with a ruler