• Consider last week's rotated factor analysis of the issue positions of political parties.
  • You can access the factor analysis here
  • Six issues loaded above .45 on the first rotated factor, called Economic Leftism
  • The "Reliability" procedure under "Scale" in the "Analyze" menu of SPSS yields this output for those six variables:

  R E L I A B I L I T Y   A N A L Y S I S   -   S C A L E   (A L P H A)
                           Mean        Std Dev       Cases
1.     BV501              .2500         2.7717        64 Govt Ownership
2.     BV502             1.6719         2.7375        64 Econ. Planning
3.     BV503             1.1250         2.4785        64 Restr. Wealth
4.     BV504             2.3438         2.6560        64 Social Welfare
5.     BV510             1.5000         2.2254        64 Nat. Integratn.
6.     BV512             2.2188         3.0104        64 Civil Rights
         
Item-total Statistics
               Scale          Scale      Corrected
               Mean         Variance       Item-            Alpha
              if Item        if Item       Total           if Item
              Deleted        Deleted    Correlation        Deleted
BV501          8.8594        81.7418        .7241           .7193
BV502          7.4375        82.8214        .7105           .7233
BV503          7.9844        86.1426        .7267           .7240
BV504          6.7656        83.0077        .7367           .7179
BV510          7.6094       109.1625        .2495           .8213
BV512          6.8906       102.3847        .2342           .8431
         
N of Cases = 64      N of Items =  6             Alpha =    .7954

• Although 121 parties were entered into the analysis, only 64 had data on all six issues.
• For these 64 cases, the reliability (alpha) of the six-item scale of "Economic Leftism" was .80 (.7954).
• The two variables BV510 and BV512 had the lowest loadings on the factor.
• This analysis shows that the alpha value would increase to .84 if BV512 were dropped.
• So the analysis was redone with only 5 items.

  R E L I A B I L I T Y   A N A L Y S I S   -   S C A L E   (A L P H A)
                           Mean        Std Dev       Cases
1.     BV501              .3059         2.8827        85 Govt Ownership
2.     BV502             1.8000         2.6222        85 Econ. Planning
3.     BV503             1.0235         2.4299        85 Restr. Wealth
4.     BV504             2.4353         2.5282        85 Social Welfare
5.     BV510             1.6235         2.1766        85 Nat. Integratn.
         
Item-total Statistics
               Scale          Scale      Corrected
               Mean         Variance       Item-            Alpha
              if Item        if Item       Total           if Item
              Deleted        Deleted    Correlation        Deleted
BV501          6.8824        54.4622        .7947           .7480
BV502          5.3882        58.1927        .7877           .7527
BV503          6.1647        62.7345        .7260           .7740
BV504          4.7529        62.8549        .6819           .7854
BV510          5.5647        83.7249        .2039           .8957
         
N of Cases =  85     N of Items =  5             Alpha =    .8330
         

• By dropping BV512, data became available for 85 parties on the remaining five issues.
• Because new cases were entered in the analysis, the alpha value was raised to .83 (not .84).
• This analysis shows that BV510 still doesn't "belong" with the others; if it were dropped from the analysis, the alpha would raise substantially.

  R E L I A B I L I T Y   A N A L Y S I S   -   S C A L E   (A L P H A)
                           Mean        Std Dev       Cases
         
1.     BV501              .2581         2.9114        93 Govt Ownership
2.     BV502             1.6452         2.6770        93 Econ. Planning
3.     BV503              .9677         2.4648        93 Restr. Wealth
4.     BV504             2.3978         2.5071        93 Social Welfare
         
Item-total Statistics
               Scale          Scale      Corrected
               Mean         Variance       Item-            Alpha
              if Item        if Item       Total           if Item
              Deleted        Deleted    Correlation        Deleted
BV501          5.0108        45.0977        .8288           .8505
BV502          3.6237        48.8894        .7994           .8607
BV503          4.3011        52.8214        .7560           .8773
BV504          2.8710        53.1571        .7259           .8872
         
N of Cases =  93    N of Items =  4              Alpha =    .8991
         

• By dropping BV510, data became available for 93 parties on the remaining four issues.
• Although the remaining scale had only four items, its reliability was raised to .90 (.8991).