|
Size: 1343
Comment:
|
← Revision 12 as of 2013-03-08 10:17:11 ⇥
Size: 1859
Comment: converted to 1.6 markup
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 3: | Line 3: |
| The spreadsheet on the previous page uses the recommended approach of Ng and Wilcox (2010) using their model 2.3 which they found controls for type I error when assumptions underlying regression analysis are relaxed. In particular it evaluates the t-statistic for the x by group interaction term over a series of bootstrap samples generated by a bootstrap macro which you need to add-in to EXCEL which you an download from [ http://www3.wabash.edu/econometrics/EconometricsBook/Basic%20Tools/ExcelAddIns/bootstrap.htm here.] Details of how to add this into SXCEL (only needs to be done once on your PC) are given on the website and in a MS Word document downlaoded from this website given [:FAQ/EXCELmed: here.] When correctly installed it should appear | The spreadsheet on the previous page uses the recommended approach of Ng and Wilcox (2010) using their model 2.3 which they found controls for type I error when assumptions underlying regression analysis are relaxed. In particular it evaluates the t-statistic for the x by group interaction term over a series of bootstrap samples generated by a bootstrap macro which you need to add-in to EXCEL which you an download from [[http://www3.wabash.edu/econometrics/EconometricsBook/Basic%20Tools/ExcelAddIns/bootstrap.htm|here.]] Details of how to add this into EXCEL (only needs to be done once on your PC) are given on the website and in a MS Word document downloaded from this website given [[FAQ/EXCELmed| here.]] When correctly installed it should appear in the top left hand corner when you select the 'Add-Ins' tab at the top of the spreadsheet. |
| Line 6: | Line 6: |
| This macro will only handle complete cases so copy and paste your complete cases into the green area of sheet 1 of the spreadsheet representing the outcome (Y), continuous variable (X) and group (C). The HC4 asymptotic t-test of Cribari-Neto (2004) should then be produced in the cells to the right of the data. Some numbers will appear in the yellow column of sheet1. This MS word file contains instructions on how to obtain the bootstrap HC4 estimate's p-value. | Once you have got the bootstrap macro read in you are ready to go. Details of how to compute the HC4 estimate and its bootstrap equivalent (estimates 2.2 and 2.3 respectively of Ng and Wilcox, 2010) are given in the MS Word document located [[attachment:bootHC4.docx|here.]] The HC4 estimate is regarded as bring more robust when assumptions underlying multiple regression are violated (Cribari-Neto, 2004) in moderation alaysis. Ng and Wilcox have subsequently suggested that the bootstrap variant of HC4 better controls type I error rate. __References__ Cribari-Neto, F. (2004). Asymptotic inference under heteroscedasticity of unknown form. ''Computational Statistics and Data Analysis'' '''45''' 215-233. Ng, M. and Wilcox, R.R. (2010). Comparing the regression slopes of independent groups. ''British Journal of Mathematical and Statistical Psychology'' '''63''' 319-340. |
Using the EXCEL spreadsheet to compute the HC4-based bootstrap estimate (estimate 2.3) of Ng and Wilcox (2010)
The spreadsheet on the previous page uses the recommended approach of Ng and Wilcox (2010) using their model 2.3 which they found controls for type I error when assumptions underlying regression analysis are relaxed. In particular it evaluates the t-statistic for the x by group interaction term over a series of bootstrap samples generated by a bootstrap macro which you need to add-in to EXCEL which you an download from here. Details of how to add this into EXCEL (only needs to be done once on your PC) are given on the website and in a MS Word document downloaded from this website given here. When correctly installed it should appear in the top left hand corner when you select the 'Add-Ins' tab at the top of the spreadsheet.
Once you have got the bootstrap macro read in you are ready to go. Details of how to compute the HC4 estimate and its bootstrap equivalent (estimates 2.2 and 2.3 respectively of Ng and Wilcox, 2010) are given in the MS Word document located here. The HC4 estimate is regarded as bring more robust when assumptions underlying multiple regression are violated (Cribari-Neto, 2004) in moderation alaysis. Ng and Wilcox have subsequently suggested that the bootstrap variant of HC4 better controls type I error rate.
References
Cribari-Neto, F. (2004). Asymptotic inference under heteroscedasticity of unknown form. Computational Statistics and Data Analysis 45 215-233.
Ng, M. and Wilcox, R.R. (2010). Comparing the regression slopes of independent groups. British Journal of Mathematical and Statistical Psychology 63 319-340.