Table 3 Results of Chi-Square test and linear regression analysis for contingency tables 25 × 17 classes.

From: Evaluation of two semi-supervised learning methods and their combination for automatic classification of bone marrow cells

  

CST

AL

CST + AL

 

N

408

408

408

Number of newly labeled teacher data by each round in each class: C

 

Pearson's test

Chi-square

17,968.7

766.061

16,506.16

 
 

Prob > ChiSq

 < 0.0001

 < 0.0001

 < 0.0001

Results of linear regression analysis using least-squares method

 

N

408

408

408

 

Factor

Method

epoch

Class

 

Mean

1.272223

1.104601

1.322059

Fitting of model

p (Prob >|t|)

 < 0.0001

 < 0.0001

0.2268

R square

F-value

p(Prob > F)

SEM

0.037217

0.014357

0.058687

0.32286

13.7454

 < 0.0001

Increasing rate of C: C in nth epoch / C in (n-1)th round

Linear regression analysis

vs CST + AL

EPLRLSM

0.033691

-0.1305

 

Linear regression analysis

LTEP

4.887

77.035

0.644

   

p (t-test)

0.1913

 < 0.0001

 

F-value

11.07754

22.6495

1.2441

 

p (Prob > F)

 < .0001

 < .0001

0.2268

 

N

408

408

408

 

Difference of C : D D = C in nth round—C in (n−1)th round

 

Pearson's test

Chi-square

8775.255

27,081.11

10,906.14

 
  

Prob > ChiSq

 < 0.0001

 < 0.0001

 < 0.0001

 
 

N

408

408

408

 

Factor

Method

epoch

Class

 

Mean

3.930686

0.319314

4.637457

Fitting of model

p (Prob >|t|)

 < 0.0001

 < 0.0001

 < 0.0001

R square

F-value

p(Prob > F )

SEM

0.151593

0.018477

0.221208

0.52295

30.8507

 < 0.0001

Increasing rate vs 25 (the first number of teacher data) in all the class: D in nth round in all classes/ 25

Linear regression analysis

vs CST + AL

EPLRLSM

0.964412

-2.64696

 

Linear regression analysis

LTEP

114.282

11.863

81.832

   

p (t-test)

 < 0.0001

 < 0.0001

 

F-value

331.4837

4.6573

32.6611

 

p (Prob > F)

 < 0.0001

 < 0.0001

 < 0.0001

  1. Prob: Probability, ChiSq: Chi-square, SEM: Standard Error of Mean, EPLRLSM: Estimated parameter by linear regression analysis with the least square method, LTEP: Logarithmic transformed estimated parameter.
Back to article page