With regard to Metaregression in SPSS 29.0, I would like to ask the following questions.
1) Should I always check for outliers and if yes how I can detect them?
2) When I create a scatter dot with dependent and independent variables of Metaregression the sizes of the dots should be different (eg. according to 1/variance?) and how I can perform this graph.
3) What are the usual parameters that I should report after a Metaregression. If the p value of Wald Chi-square test is < 0.05 and the p value of the parameter estimate is > 0.05, the association between the dependent and independent variable is significant or not?

Thank you in advance

Best regards

George Christou

------------------------------
Georgios Christou
------------------------------

#SPSSStatistics

Hi. From one of our statisticians:

1. I would say yes to examine potential outliers. One way is to investigate the residuals. /SAVE RESID can predict the residuals. A quick check is to compare the absolute residuals with 1.96. A study with a residual greater than 1.96 could be suspicious.
2.  META REGRESSION has the /BUBBLEPLOT subcommand which can create a scatter plot with studies drawn proportional to weights.
3.  The two p-values have different interpretations. The Wald Chi-squared test tests the entire meta-regression model: Whether all model parameters (not including the intercept term) are equal to zero. The p-value of the parameter could be used to draw inference about the particular predictor: Whether it is zero given the other effects in the model. What to report is kind of subjective, depending on the research purposes.

------------------------------
Rick Marcantonio
Quality Assurance
IBM
------------------------------

Original Message:
Sent: Sat December 31, 2022 03:05 AM
From: Georgios Christou
Subject: Metaregression

With regard to Metaregression in SPSS 29.0, I would like to ask the following questions.
1) Should I always check for outliers and if yes how I can detect them?
2) When I create a scatter dot with dependent and independent variables of Metaregression the sizes of the dots should be different (eg. according to 1/variance?) and how I can perform this graph.
3) What are the usual parameters that I should report after a Metaregression. If the p value of Wald Chi-square test is < 0.05 and the p value of the parameter estimate is > 0.05, the association between the dependent and independent variable is significant or not?

Thank you in advance

Best regards

George Christou

------------------------------
Georgios Christou
------------------------------

#SPSSStatistics

On follow-up, one of the statisticians asked this to be posted:

Both I-squared and H-squared (or just H) are commonly reported to summarize the impact of heterogeneity. I-squared is a transformation of H statistic that describes the proportion of total variation due to heterogeneity. I-squared is a percentage, which is bounded, and may be easier to interpret. One or both should be fine.

------------------------------
Rick Marcantonio
Quality Assurance
IBM
------------------------------

Original Message:
Sent: Tue January 03, 2023 12:59 PM
From: Rick Marcantonio
Subject: Metaregression

Hi. From one of our statisticians:

1. I would say yes to examine potential outliers. One way is to investigate the residuals. /SAVE RESID can predict the residuals. A quick check is to compare the absolute residuals with 1.96. A study with a residual greater than 1.96 could be suspicious.
2.  META REGRESSION has the /BUBBLEPLOT subcommand which can create a scatter plot with studies drawn proportional to weights.
3.  The two p-values have different interpretations. The Wald Chi-squared test tests the entire meta-regression model: Whether all model parameters (not including the intercept term) are equal to zero. The p-value of the parameter could be used to draw inference about the particular predictor: Whether it is zero given the other effects in the model. What to report is kind of subjective, depending on the research purposes.

For me, I would give

• • How the effect sizes are measured.
  • Model type
  • Weight type
  • Model coefficient test
  • Parameter estimates
  • Residual heterogeneity
  • A corresponding scatter or bubble plot

Here I also suggest a very nice article to read and follow the example given in it to report the results.

 

Viechtbauer, Wolfgang, and Mike W‐L. Cheung. "Outlier and influence diagnostics for meta‐analysis." Research synthesis methods 1.2 (2010): 112-125.

------------------------------
Rick Marcantonio
Quality Assurance
IBM

Original Message:
Sent: Sat December 31, 2022 03:05 AM
From: Georgios Christou
Subject: Metaregression

With regard to Metaregression in SPSS 29.0, I would like to ask the following questions.
1) Should I always check for outliers and if yes how I can detect them?
2) When I create a scatter dot with dependent and independent variables of Metaregression the sizes of the dots should be different (eg. according to 1/variance?) and how I can perform this graph.
3) What are the usual parameters that I should report after a Metaregression. If the p value of Wald Chi-square test is < 0.05 and the p value of the parameter estimate is > 0.05, the association between the dependent and independent variable is significant or not?

Thank you in advance

Best regards

George Christou

------------------------------
Georgios Christou
------------------------------

#SPSSStatistics