warningsdistributionalphareliability

Warning Message: In matrix(unlist(drop.item), ncol = 10, byrow = TRUE) : data length [16] is not a multiple of the number of columns [10] >?

I've been working on a data analysis with no hick-ups whatsoever. But when I try to find the Cronbach's alpha for certain(!) variable combinations - R insists that the data length is not a multiple of the number of columns.

Here is one example:

PsycLoc <- subset(SCWB,select=c(Belong_2, Belong_3))
PSYCLOC <-data.matrix(PsycLoc)
alpha(PSYCLOC, check.keys=TRUE)

Reliability analysis   
Call: alpha(x = PSYCLOC, check.keys = TRUE)

  raw_alpha std.alpha G6(smc) average_r S/N   ase mean  sd median_r
      0.69      0.69    0.53      0.53 2.2 0.017  8.1 1.8     0.53

 lower alpha upper     95% confidence boundaries
0.66 0.69 0.72 

 Reliability if an item is dropped:
         raw_alpha std.alpha G6(smc) average_r S/N alpha se var.r med.r
Belong_2      0.53      0.53    0.28      0.53  NA       NA  0.53  0.53
Belong_3      0.28      0.53      NA        NA  NA       NA  0.28  0.53

To give you an idea of data distribution, this is SCWB$Belong_2

  1     2     3     4     5     6     7     8 8.211     9    10 
 18    11    13    25    82    95   135   275    15   156   527

And this is SCWB$Belong_3

 1     2     3     4     5     6     7 7.992     8     9    10 
 20    20    19    26   105    82   159    28   269   161   463

I do receive an alpha (0.69). Is this alpha valid? Can I just ignore the warning message and continue with the data analysis? Would it mess with the integrity of a follow up Principal Component Analysis?

If the warning message IS detrimental - how can I fix this problem? Thank you for your advice

Solution

  • the message is indeed not exactly informative (related to something in the psych::alpha computation)... but you are getting this result because you are loading Cronbach alpha with just 2 items. You need to have at least 3 items...