+ Reply to Thread
Results 1 to 6 of 6

Thread: Interpretation partial correlation table

  1. #1
    Points: 480, Level: 9
    Level completed: 60%, Points required for next Level: 20

    Posts
    7
    Thanks
    2
    Thanked 0 Times in 0 Posts

    Interpretation partial correlation table




    Hi guys, I am not very skilled in statistics and spss and am for the most part only familiar with some basic analysis methods. For an educational research assignment I have to construct a similar regression model as the one below. However, I am having some troubles interpreting this table. Could someone explain to me how I read this table? Did they to a multiple regression analysis for this table? Thank you very much!


  2. #2
    Points: 2,626, Level: 31
    Level completed: 18%, Points required for next Level: 124

    Location
    Dallas, TX
    Posts
    311
    Thanks
    12
    Thanked 94 Times in 93 Posts

    Re: Interpretation partial correlation table

    Why does your title say partial correlation table? The output doesn't say that. It seems like a correlations table. Correlation is a measure of how do two variables move together 'linearly' i.e. if x increases then so does y or may be y increases in opposite direction. The correlation of the variable with itself is always 1. So 1(intention to quit) has a value of 1 against itself & similarly you can see 1 in 2-2,3-3 & so on. For 2-1 you see a value of 0.020 i.e. correlation between # of skills & intention to quit is 0.02 or 2% which is a weak but positive correlation. The correlation between 3 & 1 is -ve which means that they move in opposite direction. A star against a number indicates that the correlation is statistically significant that means there is less than 1% chance that correlation will be equal to 0.

    These variables as you see are correlated among each other. So while measuring relationship between two variables the correlation gets influenced by other variables. If you measure the correlation after removing the influence/effect of other variables then it is called partial correlation (which I suspect is not given in this table). For example 5 & 8 have a strong correlation (47%) & so do 6 & 8(45%). Therefore, the correlation between 5&6 will be heavily influenced by 8. The partial correlation between 5 & 6 would be measured after removing effect of 8 from both 5 & 6.

    Hope it helps.

  3. The Following User Says Thank You to jrai For This Useful Post:

    Julien (02-13-2012)

  4. #3
    Points: 480, Level: 9
    Level completed: 60%, Points required for next Level: 20

    Posts
    7
    Thanks
    2
    Thanked 0 Times in 0 Posts

    Re: Interpretation partial correlation table

    Thanks! Very helpful. So to verify. The correlation between 8 Gross full-time equivalent monthly wages and 6 Tenure is 0.453* which is 45%, and the correlation between 6 Tenure and 3 PSED is -113* which is -11%. This means that a change in tenure affects PSED by -11,3%?

    Another question: Can this table only be constructed by integrating multiple analysis manually or can I get such a table in one output by means of a multiple regression analysis?

  5. #4
    Points: 2,626, Level: 31
    Level completed: 18%, Points required for next Level: 124

    Location
    Dallas, TX
    Posts
    311
    Thanks
    12
    Thanked 94 Times in 93 Posts

    Re: Interpretation partial correlation table

    Quote Originally Posted by Julien View Post
    Thanks! Very helpful. So to verify. The correlation between 8 Gross full-time equivalent monthly wages and 6 Tenure is 0.453* which is 45%, and the correlation between 6 Tenure and 3 PSED is -113* which is -11%. This means that a change in tenure affects PSED by -11,3%?
    Yes that is partially correct. I say partially correct because your choice of words has a hint of causality. Correlation doesn't mean causality. 6 & 3 move together in opposite directions but then you can't say that 6 affects/ causes 3 or vice-versa. For implying causality you need to have a random allocation of subjects. Do a complete random study where you measure PSED for different tenures. Then do a regression to find the effect of tenure on PSED. Randomized experiments can be used to infer causality.

    If you want to find just the effect without implying causality then a multiple linear regression is suffice without the requirement of randomized experiment.

    Quote Originally Posted by Julien View Post
    Another question: Can this table only be constructed by integrating multiple analysis manually or can I get such a table in one output by means of a multiple regression analysis?
    I don't program in SPSS but use the point & click environment. Somebody else with knowledge of SPSS will join & tell you the programming statements for doing it. For point & click SPSS use Analyze>Regression>Linear In the window press Statistics & check descriptives & part and partial correlations. In the Linear window you can select your dependent & independents for the multiple linear regression.

  6. The Following User Says Thank You to jrai For This Useful Post:

    Julien (02-13-2012)

  7. #5
    Points: 480, Level: 9
    Level completed: 60%, Points required for next Level: 20

    Posts
    7
    Thanks
    2
    Thanked 0 Times in 0 Posts

    Re: Interpretation partial correlation table

    Thanks again, so far this is clear to me.

    My study works with a similar set of variables. I want to construct a similar table by using the following variables: 1 intention to quit 2 PSED 3 age 4 tenure and 5 number of contractual working hours.

    However, could you explain why I should not use: analyze: correlate bivariate, since you mention that is a correlation table.
    Last edited by Julien; 02-13-2012 at 09:52 AM.

  8. #6
    Points: 2,626, Level: 31
    Level completed: 18%, Points required for next Level: 124

    Location
    Dallas, TX
    Posts
    311
    Thanks
    12
    Thanked 94 Times in 93 Posts

    Re: Interpretation partial correlation table


    You can use that if you want just the correlations table. I told you the method through regression because from your question it seemed that you wanted correlations output while performing regression. If you just want the correlations without doing the regression then correlate bivariate is the simplest way.

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats