|
Size: 1361
Comment:
|
← Revision 5 as of 2013-03-08 10:17:41 ⇥
Size: 1368
Comment: converted to 1.6 markup
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 3: | Line 3: |
| This [attachment:polychoric.xls spreadsheet] may be used to evaluate polychoric correlations. The raw data with upto four categories per variable (coded 1 to 4) is entered in sheet one for upto 15 variables and a data set having upto 800 points. Polychoric correlations are then outputted in sheet 2. Note: for skewed variables leading to the possible non-occurrence of category combinations for variable pairs the estimation may not be possible and #VALUE will be returned. [:FAQ/polychoric: Other options] for estimating polychorics are available which may handle more substantial skew than this spreadsheet. | This [[attachment:polychoric.xls|spreadsheet]] may be used to evaluate polychoric correlations. The raw data with upto four categories per variable (coded 1 to 4) is entered in sheet one for upto 15 variables and a data set having upto 800 points. Polychoric correlations are then outputted in sheet 2. Note: for skewed variables leading to the possible non-occurrence of category combinations for variable pairs the estimation may not be possible and #VALUE will be returned. [[FAQ/polychoric| Other options]] for estimating polychorics are available which may handle more substantial skew than this spreadsheet. |
| Line 5: | Line 5: |
| One advantage of using a spreadsheet to compute polychoric correlations is that the correlation matrinx in sheet 2 may be copied and pasted into a SPSS spreadsheet and inputted for an exploratory factor analysis. See [http://imaging.mrc-cbu.cam.ac.uk/statswiki/StatsCourse2009 the Factor Analysis Graduate Talk 2009] for an example of how to do this. | One advantage of using a spreadsheet to compute polychoric correlations is that the correlation matrinx in sheet 2 may be copied and pasted into a SPSS spreadsheet and inputted for an exploratory factor analysis. See [[http://imaging.mrc-cbu.cam.ac.uk/statswiki/StatsCourse2009|the Factor Analysis Graduate Talk 2009]] for an example of how to do this. |
| Line 7: | Line 7: |
| Anecdotal evidence suggests Pearson correlations can be used for factor analysis is the variables have five or more categories and are approximately have a Normal distribution. The R 'hetcor' routine for estimating Polychoric correlations may also be installed into SPSS version 17 and above (see [attachment:joao.pdf here.]) | Anecdotal evidence suggests Pearson correlations can be used for factor analysis is the variables have five or more categories and are approximately have a Normal distribution. The R 'hetcor' routine for estimating Polychoric correlations may also be installed into SPSS version 17 and above (see [[attachment:joao.pdf|here.]]) |
EXCEL spreadsheet for polychoric correlation computation
This spreadsheet may be used to evaluate polychoric correlations. The raw data with upto four categories per variable (coded 1 to 4) is entered in sheet one for upto 15 variables and a data set having upto 800 points. Polychoric correlations are then outputted in sheet 2. Note: for skewed variables leading to the possible non-occurrence of category combinations for variable pairs the estimation may not be possible and #VALUE will be returned. Other options for estimating polychorics are available which may handle more substantial skew than this spreadsheet.
One advantage of using a spreadsheet to compute polychoric correlations is that the correlation matrinx in sheet 2 may be copied and pasted into a SPSS spreadsheet and inputted for an exploratory factor analysis. See the Factor Analysis Graduate Talk 2009 for an example of how to do this.
Anecdotal evidence suggests Pearson correlations can be used for factor analysis is the variables have five or more categories and are approximately have a Normal distribution. The R 'hetcor' routine for estimating Polychoric correlations may also be installed into SPSS version 17 and above (see here.)