节作者:Rebecca Vederhus, Sebastian Jentschke

From SPSS to jamovi: Analysis of frequencies

This comparison shows how a loglinear analysis is conducted in SPSS and jamovi. The SPSS test follows the description in chapter 19.9.2 in Field (2017), especially figure 19.7 and output 19.7 - 19.10. It uses the data set Cats and Dogs.sav which can be downloaded from the web page accompanying the book.
SPSS jamovi
In SPSS, you can run a loglinear analysis using: AnalyzeLoglinearModel Selection. In jamovi, this can be done using: AnalysesFrequenciesLog- Linear Regression.
SPSS_Menu_logLinear jamovi_Menu_logLinear
In SPSS, move the variables Animal, Training and Dance to the Factor(s) box. Then, mark all three variables and click Define Range. In this window, set Minimum as 0 and Maximum as 1. Click Continue. In the box called Model Building, click Enter in single step. In jamovi, move Animal, Training and Dance to Factors. Open the Model Builder window, click + Add New Block and move the three variables to this block.
SPSS_Input_logLinear_1 jamovi_Input_logLinear_1
SPSS_Input_logLinear_2
  Add another block. Mark all three variables and choose All 2 way from the drop-down menu. Then, add a third block and mark all three variables. Open the drop-down menu and click All 3 way.
  jamovi_Input_logLinear_2
  Open Model Fit and tick the box for Overall model test. Lastly, tick Likelihood ratio tests in the Model Coefficients window.
  jamovi_Input_logLinear_3
Only the results in the output tables K-Way and Higher-Order Effects and Partial Associations in SPSS are replicated in the jamovi analysis.
SPSS_Output_logLinear jamovi_Output_logLinear_1
jamovi_Output_logLinear_2
In the K-Way and Higher-Order Effects table, you can find df-values, likelihood ratio statistics and significance values when K = 1, 2 and 3. The SPSS results also contains Pearson chi-square statistics. The different rows show if any higher-order effects or one-way effects significantly affect the model fit. The Partial Associations table breaks the model into specific parts and tells us which two-way interactions that significantly affect the fit of the model. You can tell this by looking at the significance values for the different interactions and comparing them. In jamovi, the values that are found in the K-Way and Higher-Order Effects table in SPSS can be found in the Model Fit Measures and Model Comparisons tables. However, jamovi does not provide Pearson chi-square statistics and the number of iterations. The partial associations table in jamovi is called Omnibus Likelihood Ratio Tests and are presented in three separate tables (one for each model).

Output from the SPSS analysis contains a lot of tables that are not included in the jamovi analysis. In addition, the results from the parameter estimates tables differ from each other, and are therefore not included here.

The numerical values for the statistics are the same: χ² = 127.90, p < .001; χ² = 200.16, p < .001; χ² = 51.96, p < .001: χ² = 20.30, p < .001; χ² = 65.27, p < .001; χ² = 61.15, p < .001; χ² = 1.48; χ² = 13.76, p < .001; χ² = 13.75, p < .001; χ² = 8.61, p < .01.

If you wish to replicate those analyses using syntax, you can use the commands below (in jamovi, just copy to code below to Rj). Alternatively, you can download the SPSS output files and the jamovi files with the analyses from below the syntax.
HILOGLINEAR Animal(0 1) Training(0 1) Dance(0 1)
  /CRITERIA ITERATION(20) DELTA(.0)
  /PRINT=FREQ RESID ASSOCIATION ESTIM
  /DESIGN.
jmv::logLinear(
    data = data,
    factors = vars(Animal, Training, Dance),
    blocks = list(
        list("Animal", "Training", "Dance"),
        list(c("Animal", "Training"), c("Animal", "Dance"),
             c("Training", "Dance")),
        list(c("Animal", "Training", "Dance"))),
    refLevels = list(
         list(var = "Animal", ref = "Cat"),
         list(var = "Training", ref = "Food as Reward"),
         list(var = "Dance", ref = "No")),
     modelTest = TRUE,
     dev = FALSE,
     aic = FALSE,
     pseudoR2 = NULL,
     omni = TRUE)
SPSS output file containing the analyses jamovi file containing the analyses
References
Field, A. (2017). Discovering statistics using IBM SPSS statistics (5th ed.). SAGE Publications. https://edge.sagepub.com/field5e