User Tools

Site Tools


multiple_regression

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision
Previous revision
Next revisionBoth sides next revision
multiple_regression [2018/11/09 07:53] hkimscilmultiple_regression [2019/05/21 22:40] – [Why overall model is significant while IVs are not?] hkimscil
Line 150: Line 150:
 |          B    Std. Error    Beta          |  |          B    Std. Error    Beta          | 
 |  1.000    |  (Constant)    6.399    |  1.517    |      4.220    |  0.004    |  1.000    |  (Constant)    6.399    |  1.517    |      4.220    |  0.004   
-|      bankIncome  income    0.012    |  0.004    |  0.616    |  3.325    |  0.013   +|      income    0.012    |  0.004    |  0.616    |  3.325    |  0.013   
 |      bankfam    -0.545    |  0.226    |  -0.446    |  -2.406    |  0.047    |      bankfam    -0.545    |  0.226    |  -0.446    |  -2.406    |  0.047   
 | a Dependent Variable: bankbook  number of bank   ||||||| | a Dependent Variable: bankbook  number of bank   |||||||
  
-b에 대한 (coefficients) 유의도 테스트는 t-test를 이용하여 한다. 위의 표에서 . . . .  
  
 +====== Slope test ======
 +
 +b에 대한 (coefficients) 유의도 테스트는 t-test를 이용하여 한다. t-test는 기본적으로 트리트먼트효과 (독립변인효과 혹은 차이)를 랜덤에러인 standard error로 나누어서 구하므로, 위의 표에서 income에 대한 t value는 0.012/0.004; bankfam의 경우는 -0.545 / 0.226로 구할 수 있다. 
 +
 +독립변인이 하나일 경우에 구한 t 값은 해당 리그레션 모델의 F test값의 제곱근을 씌운 값이 된다. 독립변인이 둘 이상인 경우에는 독립변인 간의 상관관계가 존재하는 경우가 대다수이므로 t 값의 제곱이 꼭 F 값이 되지는 않는다.
 +
 +====== Beta coefficients ======
 +[[:beta coefficients]] 혹은 Standardized coefficients 참조 
  
 ====== e.g., ====== ====== e.g., ======
Line 321: Line 328:
 </code> </code>
  
-====== 무엇부터라는 문제 ======+====== Why overall model is significant while IVs are not====== 
 +see https://www.researchgate.net/post/Why_is_the_Multiple_regression_model_not_significant_while_simple_regression_for_the_same_variables_is_significant 
 + 
 +<code>> RSS = 3:10 #Right shoe size 
 +> LSS = rnorm(RSS, RSS, 0.1) #Left shoe size - similar to RSS 
 +> cor(LSS, RSS) #correlation ~ 0.99 
 +[1] 0.9983294 
 +>  
 +> weights = 120 + rnorm(RSS, 10*RSS, 10) 
 +>  
 +> ##Fit a joint model 
 +> m = lm(weights ~ LSS + RSS) 
 +>  
 +> ##F-value is very small, but neither LSS or RSS are significant 
 +> summary(m) 
 + 
 +Call: 
 +lm(formula = weights ~ LSS + RSS) 
 + 
 +Residuals: 
 +      1                                     7  
 + 4.6231 -4.8706  1.3063  0.9639 -1.3120 -6.1247  2.6604  
 +      8  
 + 2.7536  
 + 
 +Coefficients: 
 +            Estimate Std. Error t value Pr(>|t|)     
 +(Intercept)  103.116      4.832  21.339 4.19e-06 *** 
 +LSS          -27.546     11.952  -2.305   0.0694 .   
 +RSS           39.299     12.040   3.264   0.0223 *   
 +--- 
 +Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 
 + 
 +Residual standard error: 4.508 on 5 degrees of freedom 
 +Multiple R-squared:  0.9827, Adjusted R-squared:  0.9757  
 +F-statistic: 141.6 on 2 and 5 DF,  p-value: 3.964e-05 
 + 
 +>  
 +> ##Fitting RSS or LSS separately gives a significant result.  
 +> summary(lm(weights ~ LSS)) 
 + 
 +Call: 
 +lm(formula = weights ~ LSS) 
 + 
 +Residuals: 
 +    Min      1Q  Median      3Q     Max  
 +-11.044  -2.203  -0.422   2.774  12.369  
 + 
 +Coefficients: 
 +            Estimate Std. Error t value Pr(>|t|)     
 +(Intercept)  105.939      7.679   13.79 9.03e-06 *** 
 +LSS           11.401      1.115   10.22 5.11e-05 *** 
 +--- 
 +Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 
 + 
 +Residual standard error: 7.282 on 6 degrees of freedom 
 +Multiple R-squared:  0.9457, Adjusted R-squared:  0.9366  
 +F-statistic: 104.5 on 1 and 6 DF,  p-value: 5.113e-05 
 + 
 +>  
 +</code> 
 + 
 + 
 +====== The problem of "which one is entered first?" ======
  
 __그림 여기쯤 수록__ __그림 여기쯤 수록__
Line 339: Line 409:
     * . . . the stepwise procedure defines an a posteriori order based solely on a statistical consideration (the statistical significance of semi-partial correlations) . . . .     * . . . the stepwise procedure defines an a posteriori order based solely on a statistical consideration (the statistical significance of semi-partial correlations) . . . .
 ====== Determining IVs' role ====== ====== Determining IVs' role ======
 +For a complete explanation and examples, read [[:partial  and semipartial correlation]]
 https://www.youtube.com/watch?v=-QsMvrQDxyU https://www.youtube.com/watch?v=-QsMvrQDxyU
 [{{ :partial.correlations.jpg?300 |r-squared semi-partial partial correlations }}] [{{ :partial.correlations.jpg?300 |r-squared semi-partial partial correlations }}]
Line 396: Line 467:
   * LifeSat Score on Life Satisfaction Inventory seven years after College   * LifeSat Score on Life Satisfaction Inventory seven years after College
   * Income Income seven years after College (in thousands)   * Income Income seven years after College (in thousands)
 +
  
 ====== Resources ====== ====== Resources ======
multiple_regression.txt · Last modified: 2023/10/19 08:39 by hkimscil

Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki