MSP output does not show scalability coefficient

[Joost de Moor]

Dear colleagues,

I have tried to do a Mokken Scale Analysis with MSP in stata. All seems to work ok, except I don't get scalability coefficients for each generated scale (i.e. the Loevinger's H at the bottom of each scale). See attachment for output. I can't find anything on why this may occur, so I was hoping someone here might be able to help.

Many thanks,
Joost

[Weiwen Ng]

Joost,

The FAQ asks us not to post attachments like screenshots. You can report your output in the code delimiters (see my signature for details).

The MSP package, by Jean-Benoit Hardouin and available on Statistical Software Components, does Mokken scale analysis, a type of non-parametric item response theory.* You use the -msp- command in the package, which is an automated scaling procedure that selects items stepwise and proposes one or more scales. While the H-coefficient (this indicates how strongly the items in each scale scale together in a Guttman-like sense (this is a simplification)) isn't printed in your screen, if you type

return list

, you should see some scalars (like 1x1 vectors) called r(H1), r(H2), etc. You can ask Stata to print a scalar by typing

di r(H1)

in the command line. -di- is shorthand for display, although you can obviously type -display- if you prefer.

* Not trying to be tendentious, but this is to make things easier for other readers who have no idea what MSP is.

[Joost de Moor]

Many thanks Weiwen for your answer and apologies for not posting my output in the correct manner.

I have used the method you suggest, but exactly the same comes out after "return list" - ie output with the H coefficients left blank. Any other suggestions?

[Joost de Moor]

And here is the output (hopefully) in the correct fashion:

. msp qid12_1 - qid12_20 lifestyle_other polpart_other, c(.4)

Scale: 1
----------
Significance level: 0.000216
The two first items selected in the scale 1 are qid12_1 and qid12_3 (Hjk=0.7920)
The following items are excluded at this step: qid12_6 qid12_8 qid12_10 qid12_11 qid12_12 qid12_13 qid12_1
> 4 qid12_15 qid12_16 lifestyle_other
Significance level: 0.000207
The item qid12_2 is selected in the scale 1      Hj=0.4285      H=     .
Significance level: 0.000200
None new item can be selected in the scale 1 because all the Hj are lesser than .4 or none new item has
all the related Hjk coefficients significantly greater than 0
.

                                  Observed    Expected                                      Number
                      Easyness     Guttman     Guttman    Loevinger               H0: Hj<=0  of NS
Item          Obs      P(Xj=1)      errors      errors      H coeff     z-stat.     p-value    Hjk
---------------------------------------------------------------------------------------------------
qid12_2       2113      0.3365         281      491.66      0.42847     18.2554     0.00000      0
qid12_1       2113      0.1159         130      350.05      0.62862     23.5504     0.00000      0
qid12_3       2113      0.2347         229      516.59      0.55671     25.8566     0.00000      0
---------------------------------------------------------------------------------------------------
Scale         2113                     320      679.15                  27.3778     0.00000


Scale: 2
----------
Significance level: 0.000292
The two first items selected in the scale 2 are qid12_6 and qid12_7 (Hjk=0.6998)
The following items are excluded at this step: lifestyle_other
Significance level: 0.000267
The item qid12_5 is selected in the scale 2      Hj=0.4865      H=     .
Significance level: 0.000248
None new item can be selected in the scale 2 because all the Hj are lesser than .4 or none new item has
all the related Hjk coefficients significantly greater than 0
.

                                  Observed    Expected                                      Number
                      Easyness     Guttman     Guttman    Loevinger               H0: Hj<=0  of NS
Item          Obs      P(Xj=1)      errors      errors      H coeff     z-stat.     p-value    Hjk
---------------------------------------------------------------------------------------------------
qid12_5       2113      0.2120         228      443.99      0.48647     18.8475     0.00000      0
qid12_6       2113      0.6351          93      293.37      0.68299     16.3337     0.00000      0
qid12_7       2113      0.1685         213      410.42      0.48102     18.1653     0.00000      0
---------------------------------------------------------------------------------------------------
Scale         2113                     267      573.89                  21.7024     0.00000


Scale: 3
----------
Significance level: 0.000417
The two first items selected in the scale 3 are qid12_11 and qid12_12 (Hjk=0.6947)
Significance level: 0.000373
The item qid12_10 is selected in the scale 3     Hj=0.5197      H=     .
Significance level: 0.000340
The item qid12_13 is selected in the scale 3     Hj=0.4677      H=     .
The following items are excluded at this step: qid12_4
Significance level: 0.000316
The item qid12_14 is selected in the scale 3     Hj=0.4793      H=     .
The following items are excluded at this step: qid12_19
Significance level: 0.000299
The item qid12_15 is selected in the scale 3     Hj=0.4292      H=     .
Significance level: 0.000286
The item qid12_16 is selected in the scale 3     Hj=0.4148      H=     .
The following items are excluded at this step: qid12_18
Significance level: 0.000276
The item polpart_other is selected in the scale 3Hj=0.4044      H=     .
Significance level: 0.000269
None new item can be selected in the scale 3 because all the Hj are lesser than .4 or none new item has
all the related Hjk coefficients significantly greater than 0
.

                                  Observed    Expected                                      Number
                      Easyness     Guttman     Guttman    Loevinger               H0: Hj<=0  of NS
Item          Obs      P(Xj=1)      errors      errors      H coeff     z-stat.     p-value    Hjk
---------------------------------------------------------------------------------------------------
polpart_other 2113      0.1368         610     1024.15      0.40438     20.3808     0.00000      0
qid12_16      2113      0.2504        1038     1694.74      0.38752     26.1908     0.00000      0
qid12_15      2113      0.5712        1464     2603.72      0.43773     40.5256     0.00000      0
qid12_14      2113      0.6001        1239     2517.00      0.50775     45.8341     0.00000      0
qid12_13      2113      0.5329        1317     2643.32      0.50176     46.8422     0.00000      0
qid12_10      2113      0.4633        1627     2584.05      0.37037     33.8154     0.00000      0
qid12_11      2113      0.6058        1300     2488.36      0.47757     42.7070     0.00000      0
qid12_12      2113      0.4326        1181     2499.36      0.52748     46.8354     0.00000      0
---------------------------------------------------------------------------------------------------
Scale         2113                    4888     9027.35                  77.0023     0.00000


Scale: 4
----------
Significance level: 0.001786
The two first items selected in the scale 4 are qid12_18 and qid12_20 (Hjk=0.6943)
The following items are excluded at this step: qid12_8 lifestyle_other
Significance level: 0.001563
The item qid12_19 is selected in the scale 4     Hj=0.6568      H=     .
Significance level: 0.001429
None new item can be selected in the scale 4 because all the Hj are lesser than .4 or none new item has
all the related Hjk coefficients significantly greater than 0
.

                                  Observed    Expected                                      Number
                      Easyness     Guttman     Guttman    Loevinger               H0: Hj<=0  of NS
Item          Obs      P(Xj=1)      errors      errors      H coeff     z-stat.     p-value    Hjk
---------------------------------------------------------------------------------------------------
qid12_19      2113      0.3961         273      795.40      0.65678     35.2912     0.00000      0
qid12_18      2113      0.4321         252      776.34      0.67540     35.2254     0.00000      0
qid12_20      2113      0.2508         205      621.05      0.66991     29.9864     0.00000      0
---------------------------------------------------------------------------------------------------
Scale         2113                     365     1096.39                  41.1054     0.00000


Scale: 5
----------
Significance level: 0.005000
The two first items selected in the scale 5 are qid12_9 and qid12_17 (Hjk=0.6270)
Significance level: 0.003846
None new item can be selected in the scale 5 because all the Hj are lesser than .4 or none new item has
all the related Hjk coefficients significantly greater than 0
.

                                  Observed    Expected                                      Number
                      Easyness     Guttman     Guttman    Loevinger               H0: Hj<=0  of NS
Item          Obs      P(Xj=1)      errors      errors      H coeff     z-stat.     p-value    Hjk
---------------------------------------------------------------------------------------------------
qid12_9       2113      0.4468         162      434.29      0.62698     24.6524     0.00000      0
qid12_17      2113      0.3715         162      434.29      0.62698     24.6524     0.00000      0
---------------------------------------------------------------------------------------------------
Scale         2113                     162      434.29                  24.6524     0.00000


Scale: 6
----------
Significance level: 0.016667
None pair of items has a significantly positive Hjk coefficient.

. 
end of do-file

. return list

scalars:
                r(dim) =  6
                 r(H5) =  .
           r(nbitems5) =  2
                 r(H4) =  .
           r(nbitems4) =  3
                 r(H3) =  .
           r(nbitems3) =  8
                 r(H2) =  .
           r(nbitems2) =  3
                 r(H1) =  .
           r(nbitems1) =  3

macros:
          r(scalenum5) : "9 17"
             r(scale5) : "qid12_9 qid12_17"
          r(scalenum4) : "19 18 20"
             r(scale4) : "qid12_19 qid12_18 qid12_20"
          r(scalenum3) : "22 16 15 14 13 10 11 12"
             r(scale3) : "polpart_other qid12_16 qid12_15 qid12_14 qid12_13 qid12_10 qid12_11 qid12_12"
          r(scalenum2) : "5 6 7"
             r(scale2) : "qid12_5 qid12_6 qid12_7"
          r(scalenum1) : "2 1 3"
             r(scale1) : "qid12_2 qid12_1 qid12_3"

matrices:
          r(selection) :  1 x 22

. di r(H1)
.

[Weiwen Ng]

Huh, I didn't see that coming! I would take this up with the author of the command. I don't dispute that the scale H coefficients are missing, but the manual seems to be pretty clear that they should be returned in those scalars. I wonder if you have any observations that are missing some of the questions, which could cause the command to behave strangely. However, if I recall correctly, the command's default should be listwise deletion (I don't use the automated selection procedure, so I don't know for sure).

[Franziska Stoeckel]

Hi,

almost 2 years later and I run into the same problem...

I know it's unlikely but before opening a new topic, has anyone found a solution to the problem? I tried including pairwise deletion in my command but that didn't seem to solve the problem, i.e. I still don't get any scalability coefficients.

[Zoe Hyde]

I also had this problem.

It was resolved when I uninstalled the st0216 package, and instead installed the loevh msp and genscore packages.