我有一个包含多个潜在变量的数据集,目前正在测试使用R包lavaan的判别有效性和使用R包psy(R 3.5.1版)的收敛有效性。在测试收敛性有效性时,我遇到某些问题(请参阅下面的问题)。
例如,我有一个变量“ PS”(=政治技能),它具有四个子量表(“ PSNA”,“ PSII”,“ PSAS”,“ PSSA”),并具有以下项目:
data$PSNA <- data[c("PS_01","PS_06","PS_07","PS_10","PS_11","PS_15")]
data$PSII <- data[c("PS_02","PS_03","PS_04","PS_12")]
data$PSAS <- data[c("PS_09","PS_13","PS_14")]
data$PSSA <- data[c("PS_05","PS_08","PS_16","PS_17","PS_18")]
当我测试收敛性有效性时,我使用以下代码(最小工作示例,假定已加载数据集并正确指定了变量等):
library(psy)
psy::mtmm(data,list(c("PS_01","PS_06","PS_07","PS_10","PS_11","PS_15"),
c("PS_02","PS_03","PS_04","PS_12"),
c("PS_09","PS_13","PS_14"),
c("PS_05","PS_08","PS_16","PS_17","PS_18")))
我得到以下输出:
Item ScaleI Scale 1 Scale 2 Scale 3 Scale 4
1 PS_01 1 0.8380423 0.5144988 0.3454916 0.5717979
2 PS_06 1 0.6913118 0.6925180 0.6450783 0.5953375
3 PS_07 1 0.8000200 0.6554494 0.7370059 0.6526070
4 PS_10 1 0.6832371 0.4995184 0.3656475 0.6495132
5 PS_11 1 0.7486715 0.4499358 0.3038301 0.4842264
6 PS_15 1 0.7857609 0.7563410 0.6776251 0.8702897
25 PS_02 2 0.5293180 0.6135094 0.6476528 0.6775223
26 PS_03 2 0.6399255 0.7222210 0.7384163 0.5931270
27 PS_04 2 0.5889947 0.7802526 0.6949446 0.6927352
28 PS_12 2 0.5916436 0.6072019 0.5857657 0.6615552
41 PS_09 3 0.6054617 0.7812562 0.7680702 0.6992401
42 PS_13 3 0.6088767 0.7501376 0.8224281 0.7091506
43 PS_14 3 0.4408717 0.6913916 0.8814835 0.6511652
53 PS_05 4 0.3694269 0.6206938 0.4342520 0.5241523
54 PS_08 4 0.7241875 0.4227134 0.4683282 0.5745616
55 PS_16 4 0.7200101 0.6347283 0.5625519 0.7770008
56 PS_17 4 0.6260972 0.6865292 0.5261209 0.7490376
57 PS_18 4 0.4539620 0.6978974 0.8278951 0.4917656
其中等级1是PSNA,等级2是PSII,等级3是PSAS,等级4是PSSA。
现在我的问题是:
1)此代码是否具有测试PS量表的收敛有效性的一般权利?
2)我可以在代码中指定在输出中写“ PSNA”而不是“ Scale 1”,“ PSII”而不是“ Scale 2”等吗?这只是一个伦理问题,而不是最重要的问题。
3)如何正确解释此输出?如果我没记错的话,则属于同一维度的项目之间应该相互关联,并且与它们自己的规模相关,并且它们与来自其他维度的项目之间的相关关系应该为零或至少具有较低的数量级。通常认为0.3(或0.4)或更高的相关性表示属于同一子量表的两个项目的“有意义”相关性。以我的情况为例,例如对于1级量表(PSNA),这五个项目与该子量表显示出较高的相关性(0.68至0.83),而与其他量表的相关性始终较低,但仍然较高(项目“ PS_15”除外) )。那么这仍然可以接受吗?还是与其他子量表的相关性太高?
4)我还希望看到子量表(PSNA,PSII,PSAS,PSSA)之间的相关性,所以PSNA与PSII,PSNA与PSAS等如何实现?
非常感谢!