COMPGROUPS.NET | Search | Post Question | Groups | Stream | About | Register

### Help, factor analysis

• Email
• Follow

```I am measuring a latent variable from 7 manifest variables, One of
them is job insecurity. The items of job insecurity are loading on 2
factors. Factor 1's reliablity is more than the reliability of the
scale and Factor 2 gives this message:
a. The value is negative due to a negative average covariance among
items. This violates reliability model assumptions. You may want to
check item codings.

Should i collapse them under one factor, what should I do.
The same is happening with Commitment scale. It measures both

case, factor two gives a low reliability co-efficient so I collapsing
it into one factor.

Can someone help me with this, as I am stuck at this stage and there
is so much more to do in little time.
```
 0

See related articles to this posting

```On Wed, 13 Jan 2010 21:15:08 -0800 (PST), Sirjana
<sirjana.sodhi@gmail.com> wrote:

>I am measuring a latent variable from 7 manifest variables, One of
>them is job insecurity. The items of job insecurity are loading on 2
>factors. Factor 1's reliablity is more than the reliability of the
>scale and Factor 2 gives this message:
>a. The value is negative due to a negative average covariance among
>items. This violates reliability model assumptions. You may want to
>check item codings.
>
>Should i collapse them under one factor, what should I do.
>The same is happening with Commitment scale. It measures both

It seems *very*  likely that you have items that are
scored in different directions so that high "X" for one is
low "X"  for the other -- for both Factor 2, and for
Commitment   in the place of X.

Your factor analysis will have shown a factor structure
Assuming that it is the convenient and natural direction,
you want to REVERSE the scoring for one fraction of the
items -- either all of the ones that are positive, or all of the
ones that are negative.

You can reverse the scoring on typical scale items, scored
1 to N, by subtracting from (N+1).  For instance, for VARx
from 1-5,

compute NEWx = 6= VARx .

You should check your results to confirm that both the
original range and the new range are (here) 1 to 5.

>
>case, factor two gives a low reliability co-efficient so I collapsing
>it into one factor.
>
>Can someone help me with this, as I am stuck at this stage and there
>is so much more to do in little time.

Better start over, doing your computations right.

--
Rich Ulrich
```
 0

1 Replies
733 Views

Similar Articles

12/1/2013 3:59:31 AM
page loaded in 61492 ms. (0)