[Solved] SEM Chi-square error result. A: Change variables type to continuous. Make the threshold JASP more flexible. - Githubissues

jasp-stats / jasp-issues

This repository is solely meant for reporting of bugs, feature requests and other issues in JASP.

59 stars 29 forks source link

[Solved] SEM Chi-square error result. A: Change variables type to continuous. Make the threshold JASP more flexible. #359

Closed acescos closed 5 years ago

acescos commented 5 years ago

JASP version: Version 0.9.2
OS name and version: Windows 10
Bug description: There is an error in the output of SEM. SEM model: `

data <- n200

ce <-

'

cause =~ i1c + i2c + i3c

effect =~ i1e + i2e + i3e

effect ~ cause

'

fit <- sem (ce, data = data)

summary (fit)

` Please, see script and data files:

ce_model.txt n200.xlsx

The result produced by Jasp (on SEM lavaan) is Chi-square = 3.997, p-value = 0.857.

Screenshot:
Expected behaviour: The result produced by lavaan package on R software is Chi-square = 4.402, p-value (Chi-square) = 0.819.

Please, see the R output file Output_ce_model_lavaan_R.docx

AlexanderLyNL commented 5 years ago

Hi acescos,

Thanks for reporting this. The reason why the output of JASP differs from that of lavaan in R, is because JASP interprets variables as factors/ordinals, when they have fewer than 20 unique entries.

To get the same result, I've changed the variables type to "scale" in JASP, and now the JASP output is the same as that of lavaan in R, see the JASP file. In the next version of JASP you can change the threshold of 20 to whatever you think is reasonable. The new standard will be 10. I hope that this clarifies the difference between the two. If not, please reopen.

Cheers, Alexander