Вы находитесь на странице: 1из 59

Introduction to Statistics

Bob Conatser Irvine 210 Research Associate Biomedical Sciences conatser@oucom.ohiou.edu

Statistics - Definition
The scientific study of numerical data based on variation in nature. (Sokal and Rohlf) A set of procedures and rules for reducing large masses of data into manageable proportions allowing us to draw conclusions from those data. (McCarthy)

Basic Terms
Measurement assignment of a number to something Data collection of measurements Sample collected data Population all possible data Variable a property with respect to which data from a sample differ in some measurable way.

Types of Measurements
Ordinal rank order, (1st,2nd,3rd,etc.) Nominal categorized or labeled data (red, green, blue, male, female) Ratio (Interval) indicates order as well as magnitude. An interval scale does not include zero.

Types of Variables
Independent Variable controlled or manipulated by the researcher; causes a change in the dependent variable. (x-axis) Dependent Variable the variable being measured (y-axis) Discreet Variable has a fixed value Continuous Variable - can assume any value

Types of Statistics
Descriptive used to organize and describe a sample Inferential used to extrapolate from a sample to a larger population

Descriptive Statistics
Measures of Central Tendency
- Mean (average) - Median (middle) - Mode (most frequent)

Measures of Dispersion - variance


- standard deviation - standard error

Measures of Association - correlation


7

Descriptive Stats Central Tendency

Descriptive Stats Central Tendency

Descriptive Stats Central Tendency


30 29 28 27 26 25 24 23 22 21 20 1 2 3 4 5 6 7 8 9 10 11 12 13 Median and Mode Mean

Age

Rank

10

Descriptive Stats Dispersion

11

Descriptive Stats Dispersion


30 29 28 27 26

Standard Deviation Standard Error

35 30 25 20 15 10 5 0

Standard Deviation Standard Error

Age

25 24 23 22 21 20

Age

12

Descriptive Stats Association


Correlations Visit_1 Visit_1 Pearson Correlation Sig. (2-tailed) N Pearson Correlation Sig. (2-tailed) N 1 83 .354** .001 83 Visit_6 .354** .001 83 1 83

Highest Mastery Level Visit 1 vs Visit 6


0.3 0.25 0.2 Pearson Correlation = 0.354

Visit_6

Visit 6

**. Correlation is significant at the 0.01 level (2 il d)

0.15 0.1 0.05 0 0 0.1 0.2 0.3 0.4 0.5

Visit 1

13

Descriptive Statistics
Data can usually be characterized by a normal distribution. Central tendency is represented by the peak of the distribution. Dispersion is represented by the width of the distribution.

14

Descriptive Statistics
Normal distribution f(x) = (1/(*(2*)))*exp[-(1/2)*((x-)/)^2]
Mean Median Mode
Frequency

Standard Deviation

Measurement

15

Descriptive Statistics
Skew Distribution

Mode
Frequency

Mean Standard Deviation Median

Measurement

16

Descriptive Statistics
Standard Deviations

Frequency

= .5 = 1.0 = 1.5

Measurement

17

Inferential Statistics
Can your experiment make a statement about the general population? Two types 1. Parametric Interval or ratio measurements Continuous variables Usually assumes that data is normally distributed 2. Non-Parametric Ordinal or nominal measurements Discreet variables Makes no assumption about how data is distributed

18

Inferential Statistics
Null Hypothesis
Statistical hypotheses usually assume no relationship between variables. There is no association between eye color and eyesight. If the result of your statistical test is significant, then the original hypothesis is false and you can say that the variables in your experiment are somehow related.

19

Inferential Statistics - Error


Type I false positive, Type II false negative,

Statistical Result for Null Hypothesis Accepted Actual Null Hypothesis TRUE Correct FALSE Type II Error Rejected Type I Error Correct

Unfortunately, and cannot both have very small values. As one decreases, the other increases.

20

Inferential Statistics - Error


Type I Error

Frequency

/2

/2

Measurement

21

Inferential Statistics - Error


Type II Error
Statistical Result for Null Hypothesis Actual Null Hypothesis

/2

/2

1-

22

Statistical Test Decision Tree

23

Inferential Statistics - Power


The ability to detect a difference between two different hypotheses. Complement of the Type II error rate (1 - ). Fix (= 0.05) and then try to minimize (maximize 1 - ).

24

Inferential Statistics - Power


Power depends on: sample size standard deviation size of the difference you want to detect The sample size is usually adjusted so that power equals .8.

25

Inferential Statistics
Effect Size
Detectable difference in means / standard deviation Dimensionless ~ 0.2 small (low power) ~ 0.5 medium ~ 0.8 large (powerful test)

26

Inferential Statistics T-Test


Are the means of two groups different? Groups assumed to be normally distributed and of similar size. t, = (Y1 Y2) / [(12 - 22) / n] (equal sample sizes) Y1 and Y2 are the means of each group 1 and 2 are the standard deviations n is the number of data points in each group is the significance level (usually 0.05) is the degrees of freedom (2 * (n 1)) (Sokal & Rohlf)

27

Inferential Statistics T-Test


Compare calculated t, value with value from table. If calculated value is larger, the null hypothesis is false. (Lentner, C., 1982, Basle, Switzerland)

Geigy Scientific Tables vol. 2, CIBA-Geigy Limited,

28

T-Test Example
Group Treadmill 1 - healthy time 2 - diseased in seconds 1 1014 1 684 1 810 1 990 1 840 1 978 1 1002 1 1110 2 864 2 636 2 638 2 708 2 786 2 600 2 1320 2 750 2 594 2 750 Group1 mean 928.5 StDev 138.121 Group2 mean 764.6 StDev 213.75

Coronary Artery Disease


1200

Example data from SPSS Null Hypothesis There is no difference between healthy people and people with coronary artery disease in time spent on a treadmill.

1000

healthy diseased

Tim on Treadm (sec) e ill

800

600

400

200

29

T-Test Decision Tree

30

T-Test Example (cont.)


Independent Samples Test Levene's Test for Equality of Variances t-test for Equality of Means 95% Confidence Interval of the Difference Lower Upper -21.642 349.442 -13.398 341.198

F Treadmill time Equal variances in seconds assumed Equal variances not assumed .137

Sig. .716

t 1.873

df 16

Mean Std. Error Sig. (2-tailed) Difference Difference .080 .068 163.900 163.900 87.524 83.388

1.966 15.439

Null hypothesis is accepted because the results are not significant at the 0.05 level.

31

Non-Parametric Decision Tree

32

Non-Parametric Statistics
Makes no assumptions about the population from which the samples are selected. Used for the analysis of discreet data sets. Also used when data does not meet the assumptions for a parametric analysis (small data sets).

33

Non-Parametric Example I
Mann-Whitney Most commonly used as an alternative to the independent samples T-Test.
Test Statistics b Treadmill time in seconds 15.000 70.000 -2.222 .026 .027
a

Mann-Whitney U Wilcoxon W Z Asymp. Sig. (2-tailed) Exact Sig. [2*(1-tailed Sig.)]

a. Not corrected for ties. b. Grouping Variable: group

Note difference in results between this test and T-Test.

34

Inferential Statistics - ANOVA


ANOVA Analysis of Variance Compares the means of 3 or more groups Assumptions:
Groups relatively equal. Standard deviations similar. (Homogeneity of variance) Data normally distributed. Sampling should be randomized. Independence of errors.

Post-Hoc test

35

ANOVA - Example
12 13 15 15 17 19 20 20 Mean StDev StErr 15 15 16 17 17 21 21 23 23 23 25 26 27 30 30 31
35 30 25 20 15 10 5 0

Series1 Series2 Series3

16.38 18.13 26.88 3.114 3.091 3.182 0.794 0.747 0.626

36

Anova Decision Tree

37

ANOVA - Results
Multiple Comparisons

1.

3.
Test of Homogeneity of Variances VAR00001 Levene Statistic .001 df1 2 df2 21 Sig. .999

Dependent Variable: VAR00001 Tukey HSD Mean Difference (I-J) -1.75000 -10.50000* 1.75000 -8.75000* 10.50000* 8.75000*

(I) VAR00004 1.00 2.00 3.00

(J) VAR00004 2.00 3.00 1.00 3.00 1.00 2.00

Std. Error 1.56458 1.56458 1.56458 1.56458 1.56458 1.56458

Sig. .514 .000 .514 .000 .000 .000

*. The mean difference is significant at the .05 level.

2.
VAR00001 Sum of Squares 506.333 205.625 711.958

ANOVA

df 2 21 23

Between Groups Within Groups Total

Mean Square 253.167 9.792

F 25.855

Sig. .000

38

ANOVA Example 2
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Mean StDev 322.000 322.005 322.022 321.991 322.011 321.995 322.006 321.976 321.998 321.996 321.984 321.984 322.004 322.000 322.003 322.002 321.999 0.011 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 322.007 322.031 322.011 322.029 322.009 322.026 322.018 322.007 322.018 321.986 322.018 322.018 322.020 322.012 322.014 322.005 322.014 0.011 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 321.986 321.990 322.002 321.984 322.017 321.983 322.002 322.001 322.004 322.016 322.002 322.009 321.990 321.993 321.991 322.002 321.998 0.010 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 321.994 322.003 322.006 322.003 321.986 322.002 321.998 321.991 321.996 321.999 321.990 322.002 321.986 321.991 321.983 321.998 321.995 0.007

SPSS example Machine type vs. brake diameter 4 group ANOVA Null Hypothesis There is no difference among machines in brake diameter.

39

ANOVA Example 2

40

ANOVA Example 2 - Results


ANOVA Disc Brake Diameter (mm)

2.

Between Groups Within Groups Total

Sum of Squares .004 .006 .009

df 3 60 63

Mean Square .001 .000

F 11.748

Sig. .000

Null hypothesis is rejected because result is highly significant.


3.

1.
Test of Homogeneity of Variances Disc Brake Diameter (mm) Levene Statistic .697 df1 3 df2 60 Sig. .557

Multiple Comparisons Dependent Variable: Disc Brake Diameter (mm) Tukey HSD Mean Difference (I-J) -.0157487* .0157487* .0159803* .0188277* -.0159803* -.0188277*

(I) Machine Number 1 2

3 4

(J) Machine Number 2 1 3 4 2 2

Sig. .000 .000 .000 .000 .000 .000

*. The mean difference is significant at the .05 level.

41

Non-Parametric ANOVA Decision Tree

42

Non-Parametric ANOVA Example II


Kruskal-Wallis The Kruskal-Wallis test is a non-parametric alternative to one-way analysis of variance.
The test result (shown below) is highly significant. A post hoc test (multiple Mann-Whitney tests) would be done to determine which groups were different.
Test Statisticsa,b Chi-Square df Asymp. Sig. Brake_Dia 23.563 3 .000

a. Kruskal Wallis Test b. Grouping Variable: Machine

43

Chi-Square Test
used with categorical data two variables and two groups on both variables results indicate whether the variables are related
Figure from Geigy Scientific

Tables vol. 2

44

Chi-Square Test
Assumptions: - observations are independent - categories do not overlap - most expected counts > 5 and none < 1 Sensitive to the number of observations Spurious significant results can occur for large n.

45

Chi Squared Decision Tree

46

Chi-Square Example
A 1991 U.S. general survey of 225 people asked whether they thought their most important problem in the last 12 months was health or finances. Null hypothesis Males and females will respond the same to the survey.

1 - males 2 - females 2 2 2 2 2 1 2 2

1 - health 2 - finances 1 1 1 1 2 2 2 2

47

Chi-Square Example
problem * group Crosstabulation

Cross-tabulation table shows how many people are in each category. The nonsignificant result signifies that the null hypothesis is accepted.

Count group Males problem Total Health Finance 35 56 91


Chi-Square Tests Value .372b .223 .373 .371 225 df 1 1 1 1 Asymp. Sig. (2-sided) .542 .637 .541 .543 Exact Sig. (2-sided) Exact Sig. (1-sided)

Females 57 77 134

Total 92 133 225

Pearson Chi-Square a Continuity Correction Likelihood Ratio Fisher's Exact Test Linear-by-Linear Association N of Valid Cases

.582

.319

a. Computed only for a 2x2 table b. 0 cells (.0%) have expected count less than 5. The minimum expected count is 37. 21.

48

Chi-Square Example II
Chi-square test can be extended to multiple responses for two groups.
problem * group Crosstabulation Count group Males Females problem Health 35 57 Finance 56 77 Family 15 33 Personal 9 10 Miscellaneou 15 25 Total 130 202 Total 92 133 48 19 40 332
Chi-Square Tests Value 2.377a 2.400 .021 332 Asymp. Sig. (2-sided) df 4 .667 4 .663 1 .885

Pearson Chi-Squa Likelihood Ratio Linear-by-Linear Association N of Valid Cases

a. 0 cells (.0%) have expected count less than 5. T minimum expected count is 7.44.

49

50

Multiple Regression
Null Hypothesis GPA at the end of Freshman year cannot be predicted by performance on college entrance exams. GPA = * ACT score + * Read. Comp. score

51

Multiple Regression
3 2.5 2 y = 0.0382x + 1.56 R2 = 0.0981 3 2.5
2.8

y = 0.0491x + 1.2803 R2 = 0.2019

2.6 2.4

GPA

GPA

1.5 1 0.5 0 5 10 15 20

GPA

1.5 1 0.5 0 5 10 15 20

2.2 2.0 1.8 1.6 1.4 1.2 18 16 14 14 12 10

. ad Re

16

p. m Co
12

ACT Score

Reading Comprehension Score

10

6 4

re co S CT A
8

e or Sc

52

Multiple Regression
ANOVAb

The analysis shows no significant relationship between college entrance tests and GPA.

Model 1

Regression Residual Total

Sum of Squares .373 1.317 1.689

df 2 12 14

Mean Square .186 .110

F 1.699

Sig. .224a

a. Predictors: (Constant), Reading Comprehension score, ACT score b. Dependent Variable: Grade Point Average in first year of college

Coefficientsa Unstandardized Coefficients B Std. Error 1.172 .460 .018 .034 .042 .031 Standardized Coefficients Beta

Model 1

(Constant) ACT score Reading Comprehension score

= .151 = .386

t 2.551 .537 1.374

Sig. .025 .601 .195

a. Dependent Variable: Grade Point Average in first year of college

53

MANOVA
Multivariate ANalysis of VAriance (MANOVA) MANOVA allows you to look at differences between variables as well as group differences. assumptions are the same as ANOVA additional condition of multivariate normality also assumes equal covariance matrices (standard deviations between variables should be similar).

54

MANOVA Example
Subset of plantar fasciitis dataset. Null Hypothesis There is no difference in soleus emg activity, peak torque, or time to peak torque for quick stretch measurements in people with plantar fasciitis who receive counterstrain treatment compared with the same group of people receiving a placebo treatment.
Treatment Group 1 - Treated 2 - Control 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 Treated Mean StDev Control Mean StDev Peak to Peak Soleus EMG Response millivolt*2 0.0706 0.0189 0.0062 0.0396 0.0668 0.2524 0.0183 0.0393 0.1319 0.3781 0.039 0.074 0.0396 0.0143 0.076 0.2213 0.0196 0.0498 0.155 0.2887 Peak Quick Time to Stretch Peak Torque Torque milliseconds newton-meter 0.883 0.347 0.388 1.104 3.167 2.628 0.346 1.535 3.282 3.622 0.557 0.525 1.400 0.183 3.074 3.073 0.271 2.278 3.556 3.106 322.56 329.28 319.2 325.92 315.84 248.64 336 332.64 292.32 278.88 299.04 372.96 362.88 295.68 322.56 258.72 346.08 302.4 309.12 292.32 310.128 28.15859087 316.176 35.22039409

0.10221 1.73017 0.12151083 1.31802532 0.09773 1.80212 0.09324085 1.35575411

55

MANOVA Example
Plantar Study Soleus Peak to Peak EMG
0.25 0.2

Plantar Study Peak Torque


3.5 3

Treated Control

Plantar Study Time to Peak Torque


Treated Control
400 350 300

Treated Control

0.15 0.1

T o rq u e (n m )

E M G (m v* 2 )

2.5 2

1.5 1

0.05 0

0.5 0

T im e ( m s )

250 200 150 100 50 0

-0.05

56

MANOVA Results
a Box's Test of Equality of Covariance Matrices

Box's M F df1 df2 Sig.

5.165 .703 6 2347.472 .647

Tests the null hypothesis that the observed covariance matrices of the dependent variables are equal across groups. a. Design: Intercept+group

Boxs test checks for equal covariance matrices. A non-significant result means the assumption holds true. Levenes tests checks for univariate normality. A non-significant result means the assumption holds true.

a Levene's Test of Equality of Error Variances

F Soleus peak to peak emg (mv*2) Peak quick stretch torque (nm) Time to peak torque (milliseconds) .349 .078 .550

df1 1 1 1

df2 18 18 18

Sig. .562 .783 .468

Tests the null hypothesis that the error variance of the dependent v is equal across groups. a. Design: Intercept+group

57

MANOVA Results
Multivariate Tests b Effect Intercept Pillai's Trace Wilks' Lambda Hotelling's Trace Roy's Largest Root Pillai's Trace Wilks' Lambda Hotelling's Trace Roy's Largest Root Value .996 .004 226.499 226.499 .020 .980 .020 .020 F 1207.992a 1207.992a 1207.992a 1207.992a .109a .109a .109a .109a Hypothesis df 3.000 3.000 3.000 3.000 3.000 3.000 3.000 3.000 Error df 16.000 16.000 16.000 16.000 16.000 16.000 16.000 16.000 Sig. .000 .000 .000 .000 .954 .954 .954 .954

group

a. Exact statistic b. Design: Intercept+group

The non-significant group result indicates that the null hypothesis is true. If the result had been significant, you would need to do post hoc tests to find out which variables were significant.

58

References
Bruning, James L. and Kintz, B.L.Computational handbook of statistics (4th Edition). Massachusetts: Addison Wesley Longman, Inc., 1997.

Research Resource Guide 3rd Edition 2004-2005, http://ohiocoreonline.org, go to Research->Research Education.


Norusis, Marija J. SPSS 9.0 Guide to data analysis.New Jersey: Prentice-Hall, Inc., 1999. Sokal, Robert R. and Rohlf, James F. Biometry (2nd Edition). New York: W.H. Freeman, 1981. Stevens, James. Applied Multivariate Statistics for the Social Sciences. New Jersey: Lawrence Erlbaum Associates, Inc., 1986. Lentner, Cornelius. Geigy Scientific Tables vol. 2. Basle, Switzerland: Ciba-Geigy Limited, 1982.

59

Оценить