Hi. I spoke with one of our statisticians and he replied:
There are quite a few reasons that may cause the discrepancy. I suggest that the user first check if the following settings are matched between SPSS and excel.
- Random or fixed model. Random model is the setting by default.
- Underlying distribution T or Normal. SPSS uses the t-distribution by default.
- Way to estimate the parameters. SPSS uses REML by default.
Also, I need a bit more explanations on "SPSS gave me were spot on and almost perfect." Not quite clear what this means.
------------------------------
Rick Marcantonio
Quality Assurance
IBM
------------------------------
Original Message:
Sent: Thu June 30, 2022 05:58 AM
From: Pratham Simaria
Subject: Values Calculated from parameter estimates not matching the predicted values put in by SPSS
Hello! I used Meta-Regression in Meta Analysis. I set my target variable and opted for the predicted values to be displayed. After running it, I took out the parameter estimates (which I assumed are coefficients) as I wanted to predict future values, and calculated the values in excel just to confirm. The values I got in excel were a bit different than the predicted values I got from SPSS. They are not very far off but the predicted values that SPSS gave me were spot on and almost perfect. I tried out different decimal points and re-checked everything. How do I get the same values from the parameter estimates?
------------------------------
Pratham Simaria
------------------------------
#SPSSStatistics