If both Cronbach's alpha and McDonald's omega are above .7, the scale "by definition" is considered acceptable in terms of reliability. Removing an item solely to increase these statistics might not be necessary if the scale is functioning well for its intended purpose. But did you perform a factor analysis first? That's usually a good starting point to check how well your items measure the constructs before looking into reliability.

Were the two items that you removed genuinely negatively correlated with the others, or were they just reverse-worded? Coefficient alpha assumes that all the items are positively correlated, so if your have an item that is scored in the opposite direction as the others, then you need to recode that item.

So first, check the wording on the two items that you removed and recode them If they do indeed belong in the data set. That might raise your alpha to a level where you don't need to worry about which items to include.

But if you still get some items with lower item-to-total correlations, than you may indeed want to try factor analysis. FYI my own approach is to start with alpha to see if the items fit a single-factor model, and only proceed to actual factor analysis if there is any indication of multiple factors.

interitem correlation less than 0.400 are problematic. Removing them can enahcne the reliablity of the construct

Deciding whether to remove an item from a scale to improve reliability coefficients like Cronbach's alpha and McDonald's omega involves balancing several factors, including the theoretical importance of the item, the overall reliability of the scale, and the implications of having fewer items. Here are some considerations to help you make an informed decision:

Reliability Coefficients

• Cronbach's Alpha and McDonald's Omega:Both coefficients measure the internal consistency of the scale. Values above 0.7 are generally considered acceptable, while values above 0.8 are preferable. Removing items can sometimes increase these coefficients, but this doesn't always mean the scale is better. It's crucial to consider why an item is being removed.

Item-Rest Correlation

• Low Item-Rest Correlation:An item-rest correlation of 0.16 is relatively low, suggesting that item 7 does not correlate well with the rest of the scale. Items with low item-rest correlations may not contribute meaningfully to the construct being measured and can sometimes decrease the overall reliability.

Theoretical and Practical Considerations

• Theoretical Importance:Consider whether item 7 is theoretically important for the construct you are measuring. If it captures an essential aspect of the construct, removing it might harm the content validity of the scale. If the item is not essential, removing it may be justified to improve the reliability coefficients.
• Number of Items:Having only 6 items instead of the original 10 can be acceptable as long as the scale still adequately covers the construct and maintains good psychometric properties. Fewer items can sometimes make a scale more practical and easier to administer, but it's important to ensure that the scale still provides a comprehensive measure of the construct.

Specific Advice for Your Situation

• Current Situation:You already removed items 8, 9, and 10, resulting in acceptable reliability coefficients. Removing item 7 might further improve reliability, but consider the impact on the scale's content validity and the overall measurement of the construct.
• Decision to Remove Item 7:If removing item 7 significantly improves Cronbach's alpha and McDonald's omega and the item is not critical for measuring the construct, it might be worth removing it. However, if item 7 is important for the construct, you might consider keeping it, even if it has a low item-rest correlation.
• Remaining with 6 Items:Six items can still be sufficient, especially if they cover the key aspects of the construct and the reliability is good. Ensure the remaining items collectively provide a comprehensive and accurate measure of the construct.

Conclusion

Given that your goal is to see the correlation between three scales and this is one of them, maintaining good reliability is crucial. However, do not compromise the content validity of the scale by removing theoretically important items. Here’s a step-by-step approach you can consider:

By carefully considering these factors, you can make a well-informed decision that balances reliability, validity, and the practical aspects of your scale.

Pandia Vadivu Pandian When you cut and paste an answer from an AI, you should state your source, just as you would in any other academic setting.

• If Cronbach's alpha and McDonald's omega are both over 0.7, the scale is generally considered reliable. Removing an item just to increase these metrics slightly might not be necessary and could even harm the scale's content validity. It’s important to consider the item's theoretical relevance and content coverage before deciding to remove it. If the item is essential to the construct you’re measuring, it's better to keep it, even if the removal could slightly increase the reliability coefficients.

In this situation, as the two items are negatively correlated with the scale, please cross check if these are reverse coded items (check reliability after reverse coding). However, if these two items are not reverse coded, then surely you have to see the issue in data and remove the item if data seems okay.