MXCELLS in version 28 isn't supported anymore. How can I run my regression without it?

MXCELLS controlled the maximum number of cells in a pivot table. I can't imagine a regression table that would be too big to appear without increasing the limit. I don't know why this setting was removed, but my guess is that there is no practical size limit.

What sort of regression table are you trying to produce?

I am doing a very large regression analysis utilizing 250k transactions but needing 4 to 5 thousand variables. With previous versions I would just use MXCELLS to do the job, but now that no longer works so I will either have to use an older version of SPSS or switch to a different product like Python or R.

I don't know why this setting was removed. It was originally added, IIRC, to protect against users inadvertently generating gigantic tables that would take forever or would fail to render in the Viewer.

I don't know where that limitation is enforced - frontend or backend, but one thing to try if you are not already doing this would be to use OMS to capture the regression table and save it as a dataset or other format and specify VIEWER=NO to prevent it from appearing in the Viewer.

Sadly, I tried your suggestion, but it must be something besides the viewer display. Once you exceed 2 million, the execution of the command stops.

Sorry. It was worth a try. Development will look into the issue.

Development came up with this solution. The first part is just generating test data. It uses the OUTPUTSET feature of OMS. This examples writes to Excel, but other formats should also work.

input program.

vector v(8000).

loop #i=1 to 100000.

loop #j=1 to 8000.

compute v(#j)=normal(1).

end loop.

end case.

end loop.

end file.

end input program.

execute.

oms /select all except=warnings /destination viewer=no format=xls outputset=subtypes folder="C:\Temp2" /tag="r".

REGRESSION /MISSING LISTWISE

 /STATISTICS COEFF R ANOVA /CRITERIA=PIN(.05) POUT(.10)

 /NOORIGIN /DEPENDENT v1 /METHOD=ENTER v2 to v7001.

omsend tag='r'.

I tried that exactly as it was written (except for the path) and I still got the same warning message. Just to be clear, I only brought over the OMS code.

Would you post your full regression syntax? Development would like to track this down.

We are working to restore MXCELLS, which was unintentionally and mistakenly disabled in a recent release.

Just to confirm - you are running version 28?

Yes I am.

OK, thank you. Can I see your REGRESSION syntax, please? I need to know exactly which table(s) is(are) causing the problem. Thanks in advance.

oms select tables

 /destination format = sav numbered = 'table_number' outfile = !PartOfConstructionDataBeogradCentar + "RegressionSummary2_1.sav"

 /if commands = ['regression'] subtypes = ['Model Summary']

 /tag = "model_summary".

oms select tables

 /destination format = sav numbered = 'table_number' outfile = !PartOfConstructionDataBeogradCentar + "RegressionCoeficients2_1.sav"

 /if commands = ['regression'] subtypes = ['Coefficients','Excluded Variables']

 /tag = "reg".

 * ProtivPozarniSistem_BLDG SistemZaKlimatizaciju_BLDG - these want to come in negative so will remove.

SET MXCELLS 6000000.

REGRESSION

  /VARIABLES = LN_Price 

  PrvaProdaja LegalEntities Hipoteka Nelegalan_objekat Drugi_tereti NedovrsenObjekat

  LN_StanUZgradiApartman_Ratio1 LN_StanUZgradiApartman_Ratio2 LN_StanUZgradiApartman23 LN_SMALL_APT 

  LN_DrugiDeoStambeneZgrade LN_GaraznaProstorija LN_GaraznoMesto LN_ParkingMestoVanObjekta

  LN_SpratNovi1 

  Potkrovlje Medjusprat PovuceniSprat Suteren Prizemlje PodrumSprat NiskoPrizemlje VisokoPrizemlje Mansarda

  Lamela FreeStanding Solitaire ResidentialComplex

  GrejanjeSifra_central GrejanjeSifra_other

  Lift_BLDG 

  Врста_крова_Кос_BLDG Врста_крова_Раван_BLDG

  Garsonjera

  LN_PCT_GOOD

  Blizina_Aerodroma Blizina_Ambulante Blizina_ABStajalista Blizina_Autoputa Blizina_Biblioteke Blizina_Bioskopa

  Blizina_Dalekovoda Blizina_SveElelktrane Blizina_Trafostanice Blizina_DZStare Blizina_Dom_Zdravlja Blizina_Fakulteti Blizina_Hotela

  Blizina_Ind_Pos_Obj Blizina_Instituta Blizina_Jezera Blizina_Klinike Blizina_Manastira Blizina_Muzeja Blizina_OsnSkola

  Blizina_OZ_Ustanove Blizina_Parka Blizina_Pijace Blizina_Pozorista Blizina_PSko_Ustanove Blizina_SpecBolnica Blizina_SpomeniciKulture

  Blizina_SportRekCentri Blizina_SrednjaSkola Blizina_Supermarket Blizina_TaxiS Blizina_TrzniC Blizina_VaznijeSaobracajnice 

  Blizina_VelikiVodotokovi Blizina_VerskiO Blizina_VisokoSkolskeUstanove Blizina_ZeleznickeStanice VazniPutevi50m

  KatastarskaOpstina704008 KatastarskaOpstina704059

  Pva_Zona_S Poslovna_Z Druga_Z Treca_Zona Cetvrta_Zo Ekstra_Zon Zelena_Zon

  PARCID_704008_00257_001 TO PARCID_716090_06871_000

  /CRITERIA=PIN(.10) POUT (.20)

  /DESCRIPTIVES=MEAN STDDEV

 /STATISTICS COEFF OUTS CI(95) R ANOVA 

  /DEPENDENT=LN_Price

  /METHOD=BACKWARD

  /SAVE PRED(PRED) ADJPRED(ADJPRED) RESID ZRESID ZPRED COOK SRESID ICIN MCIN.

omsend.

OK, thanks. Could you try removing OUTS and, if possible, could you change to /METHOD=FORWARD?

I will give that a try and then respond once it has run.

I made the changes and everything ran without any warnings.

Hey, that's great! Very glad to hear it.

What does OUTS do? I get FORWARD, but what is OUTS?

OUTS. Statistics for variables not yet in the equation that have been named on METHOD subcommands for the equation. OUTS displays the Excluded Variables table showing beta, t, two-tailed probability of t, and minimum tolerance of the variable if it were the only variable entered next.

****

Often, when a regression has a large number of potential variables (thousands), I find this table overly large and, frankly, not altogether useful.

Thanks

Sure.

Hi

I'm having a problem running on SPSS 28 due to the loss of MXCELLS. I notice you say above (26/04/2022) you're trying to resore MXCELLS, but my recent installation still has the problem. I removed the OUTS command as suggested above, but need METHOD=ENTER as writing a research publication. I am running a hierarchical regression with 4 levels and 37 IVs, using 4000 wild bootstrap samples. Any advice would be welcome - thanks