0

I have a sample dataset of six variables each with 20 data points. Each data point is a measurement on one of the 20 subjects at one of the six points in time.

I generated a simple MANOVA report in JMP and was curious about how the partial covariance matrix was generated.

Digging through the literature I have found resources that explain how to generate these entries, but I have yet to be able to match JMP's output.

It seems like a fairly standard method, but I can't figure it out.

Here is the unbiased covariance matrix of the data:

enter image description here

Here is the partial covariance matrix of the data from JMP:

enter image description here

Does anyone know how JMP is generating their partial covariance matrix? Thank you

Raw Data:

April-AM April-PM May-AM May-PM June-AM June-PM
278      280      204    208    171     175
278      281      195    199    185     189
276      280      213    219    179     181
276      281      201    211    183     188
279      285      188    192    170     174
266      270      220    224    180     184
280      284      228    232    200     204
284      288      233    237    175     179
273      277      215    219    202     207
281      285      237    241    199     205
278      282      273    277    281     285
273      277      274    278    280     282
282      285      276    281    279     282
274      277      285    289    274     278
277      281      279    284    284     285
279      283      278    284    268     272
277      279      291    291    280     285
275      279      280    283    281     283
276      282      277    282    274     279
282      286      281    285    282     285
pj2452
  • 155
  • 1
  • 8
  • Following [this](https://stats.stackexchange.com/a/213122/3277), may you get what you observe? or not? – ttnphns Aug 17 '17 at 06:45
  • Thank you for the post, but unfortunately I could not match the JMP partial covariance output using either of those methods. – pj2452 Aug 21 '17 at 04:36
  • You should probably give your data (not just matrices). – ttnphns Aug 21 '17 at 07:25
  • I managed to add the raw dataset to original post. – pj2452 Aug 21 '17 at 19:33
  • How did you check that JMP is using exactly the same data values as your other software? – whuber Aug 21 '17 at 20:03
  • The original dataset is in JMP and I copied the values into excel and Matlab. They are all integers so I don't think there's a floating point precision issue. – pj2452 Aug 21 '17 at 21:28

0 Answers0