Question on linear regression with multiple conditions

[Goh jedrek]

I am supposed to generate a linear regression with multiple conditions.

Data is as shown below:

input byte id float time int conc str8 diet float(elim pred lnalcohol pred_rep)
1    3  75 "noGuar"   .31319055 4.3103285  4.317488 4.4182615
1  3.5  64 "noGuar"   .31319055 4.1537333  4.158883 4.2178164
1    4  53 "noGuar"   .31319055  3.997138  3.970292 4.0173717
1  4.5  47 "noGuar"   .31319055  3.840543 3.8501475  3.816927
1    5  40 "noGuar"   .31319055 3.6839476 3.6888795  3.616482
1  3.5  65 "withGuar" .28823623   4.18296 4.1743875 4.2178164
1    4  57 "withGuar" .28823623 4.0388417 4.0430512 4.0173717
1  4.5  50 "withGuar" .28823623  3.894724  3.912023  3.816927
1    5  42 "withGuar" .28823623 3.7506056   3.73767  3.616482
2    3  82 "noGuar"    .4008896 4.4182615  4.406719 4.4182615
2  3.5  68 "noGuar"    .4008896 4.2178164 4.2195077 4.2178164
2    4  57 "noGuar"    .4008896 4.0173717 4.0430512 4.0173717
2  4.5  45 "noGuar"    .4008896  3.816927 3.8066626  3.816927
2    5  37 "noGuar"    .4008896  3.616482  3.610918  3.616482
2 1.33 122 "withGuar" .22233166  4.846237  4.804021  5.087747
2 1.67 124 "withGuar" .22233166  4.770644  4.820282  4.951445
2    2 108 "withGuar" .22233166  4.697274 4.6821313  4.819151
2  2.5 103 "withGuar" .22233166 4.5861087  4.634729  4.618706
2    3  92 "withGuar" .22233166 4.4749427 4.5217886 4.4182615
2  3.5  65 "withGuar" .22233166 4.3637767 4.1743875 4.2178164
2    4  75 "withGuar" .22233166  4.252611  4.317488 4.0173717
2  4.5  68 "withGuar" .22233166  4.141445 4.2195077  3.816927
2    5  54 "withGuar" .22233166  4.030279  3.988984  3.616482
3    4  17 "noGuar"   1.2237754    2.8603  2.833213 4.0173717
3  4.5  10 "noGuar"   1.2237754 2.2484121 2.3025851  3.816927
3    5   5 "noGuar"   1.2237754 1.6365244  1.609438  3.616482
3    4  26 "withGuar"  .9555115 3.2947736 3.2580965 4.0173717
3  4.5  18 "withGuar"  .9555115  2.817018  2.890372  3.816927
3    5  10 "withGuar"  .9555115  2.339262 2.3025851  3.616482
4  3.5  37 "noGuar"   .57818496  3.668399  3.610918 4.2178164
4    4  30 "noGuar"   .57818496 3.3793066 3.4011974 4.0173717
4  4.5  25 "noGuar"   .57818496  3.090214  3.218876  3.816927
4    5  15 "noGuar"   .57818496  2.801122   2.70805  3.616482
4  2.5  60 "withGuar"  .3793953 4.0677114 4.0943446  4.618706
4    3  46 "withGuar"  .3793953  3.878014 3.8286414 4.4182615
4  3.5  40 "withGuar"  .3793953  3.688316 3.6888795 4.2178164
4    4  35 "withGuar"  .3793953 3.4986186  3.555348 4.0173717
4  4.5  26 "withGuar"  .3793953  3.308921 3.2580965  3.816927
4    5  23 "withGuar"  .3793953  3.119223  3.135494  3.616482
5    4  20 "noGuar"    1.609438  3.032923  2.995732 4.0173717
5  4.5  10 "noGuar"    1.609438  2.228204 2.3025851  3.816927
5    5   4 "noGuar"    1.609438  1.423485 1.3862944  3.616482
5    4  22 "withGuar"  1.299283  3.105544 3.0910425 4.0173717
5  4.5  12 "withGuar"  1.299283  2.455903  2.484907  3.816927
5    5   6 "withGuar"  1.299283 1.8062614 1.7917595  3.616482
6    4  25 "noGuar"    .9162908 3.2620895  3.218876 4.0173717
6  4.5  18 "noGuar"    .9162908  2.803944  2.890372  3.816927
6    5  10 "noGuar"    .9162908  2.345799 2.3025851  3.616482
6 1.67  50 "withGuar"  .3639005  3.892754  3.912023  4.951445
6    2  42 "withGuar"  .3639005  3.772667   3.73767  4.819151
6  2.5  32 "withGuar"  .3639005  3.590717  3.465736  4.618706
6    3  35 "withGuar"  .3639005 3.4087665  3.555348 4.4182615
6  3.5  26 "withGuar"  .3639005  3.226816 3.2580965 4.2178164
6    4  23 "withGuar"  .3639005  3.044866  3.135494 4.0173717
6  4.5  15 "withGuar"  .3639005  2.862916   2.70805  3.816927
6    5  15 "withGuar"  .3639005 2.6809654   2.70805  3.616482
7    4  40 "noGuar"    .3566749  3.703815 3.6888795 4.0173717
7  4.5  35 "noGuar"    .3566749 3.5254774  3.555348  3.816927
7    5  28 "noGuar"    .3566749   3.34714 3.3322046  3.616482
7    3  61 "withGuar"  .1965659  4.101937 4.1108737 4.4182615
7  3.5  55 "withGuar"  .1965659  4.003654 4.0073333 4.2178164
7    4  48 "withGuar"  .1965659  3.905371  3.871201 4.0173717
7  4.5  46 "withGuar"  .1965659  3.807088 3.8286414  3.816927
end

My current code:

regress lnalcohol time if diet == "withGuar" & id == 1 
regress lnalcohol time if diet == "noGuar" & id == 1
regress lnalcohol time if diet == "withGuar" & id == 2
regress lnalcohol time if diet == "noGuar" & id == 2

predict pred_rep, xb

I am supposed to verify pred, with the new variable, pred_rep which is similar to pred. To do this, I need to create a regression line. The current problem is that I only get the correct pred_rep values for the last line of code. In this case,

regress lnalcohol time if diet == "noGuar" & id == 2

gives me the correct output for pred_rep but not for the other diet/id combinations.

How do I combine the regression codes above to form one output to resolve this problem?

[Joseph Coveney]

I am supposed to verify pred, with the new variable . . .
How do I combine the regression codes above to form one output to resolve this problem?

Something like the following should work.

.ÿ
.ÿversionÿ17.0

.ÿ
.ÿclearÿ*

.ÿ
.ÿquietlyÿinputÿbyteÿidÿfloatÿtimeÿintÿconcÿstr8ÿdietÿ///
>ÿÿÿÿÿÿÿÿÿfloat(elimÿpredÿlnalcoholÿpred_rep)

.ÿ
.ÿlabelÿdefineÿGuarStatusÿ0ÿnoGuarÿ1ÿwithGuar

.ÿencodeÿdiet,ÿgenerate(dit)ÿlabel(GuarStatus)ÿnoextend

.ÿ
.ÿregressÿlnalcoholÿc.time##i.dit##i.idÿifÿinlist(id,ÿ1,ÿ2)

ÿÿÿÿÿÿSourceÿ|ÿÿÿÿÿÿÿSSÿÿÿÿÿÿÿÿÿÿÿdfÿÿÿÿÿÿÿMSÿÿÿÿÿÿNumberÿofÿobsÿÿÿ=ÿÿÿÿÿÿÿÿ23
-------------+----------------------------------ÿÿÿF(7,ÿ15)ÿÿÿÿÿÿÿÿ=ÿÿÿÿÿ94.19
ÿÿÿÿÿÿÿModelÿ|ÿÿ2.60723948ÿÿÿÿÿÿÿÿÿ7ÿÿ.372462783ÿÿÿProbÿ>ÿFÿÿÿÿÿÿÿÿ=ÿÿÿÿ0.0000
ÿÿÿÿResidualÿ|ÿÿÿ.05931505ÿÿÿÿÿÿÿÿ15ÿÿ.003954337ÿÿÿR-squaredÿÿÿÿÿÿÿ=ÿÿÿÿ0.9778
-------------+----------------------------------ÿÿÿAdjÿR-squaredÿÿÿ=ÿÿÿÿ0.9674
ÿÿÿÿÿÿÿTotalÿ|ÿÿ2.66655453ÿÿÿÿÿÿÿÿ22ÿÿ.121207024ÿÿÿRootÿMSEÿÿÿÿÿÿÿÿ=ÿÿÿÿ.06288

-------------------------------------------------------------------------------
ÿÿÿÿlnalcoholÿ|ÿCoefficientÿÿStd.ÿerr.ÿÿÿÿÿÿtÿÿÿÿP>|t|ÿÿÿÿÿ[95%ÿconf.ÿinterval]
--------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿtimeÿ|ÿÿ-.3131906ÿÿÿÿ.039771ÿÿÿÿ-7.87ÿÿÿ0.000ÿÿÿÿ-.3979605ÿÿÿ-.2284207
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿÿÿÿditÿ|
ÿÿÿÿwithGuarÿÿ|ÿÿÿ-.058114ÿÿÿ.2902195ÿÿÿÿ-0.20ÿÿÿ0.844ÿÿÿÿ-.6767022ÿÿÿÿ.5604743
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿdit#c.timeÿ|
ÿÿÿÿwithGuarÿÿ|ÿÿÿ.0249545ÿÿÿ.0688854ÿÿÿÿÿ0.36ÿÿÿ0.722ÿÿÿÿ-.1218714ÿÿÿÿ.1717803
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿÿÿ2.idÿ|ÿÿÿ.3710292ÿÿÿ.2284672ÿÿÿÿÿ1.62ÿÿÿ0.125ÿÿÿÿÿ-.115937ÿÿÿÿ.8579955
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿid#c.timeÿ|
ÿÿÿÿÿÿÿÿÿÿÿ2ÿÿ|ÿÿ-.0876989ÿÿÿ.0562447ÿÿÿÿ-1.56ÿÿÿ0.140ÿÿÿÿ-.2075817ÿÿÿÿ.0321839
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿdit#idÿ|
ÿÿwithGuar#2ÿÿ|ÿÿÿ-.420878ÿÿÿÿ.336965ÿÿÿÿ-1.25ÿÿÿ0.231ÿÿÿÿ-1.139102ÿÿÿÿÿ.297346
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
dit#id#c.timeÿ|
ÿÿwithGuar#2ÿÿ|ÿÿÿ.1536034ÿÿÿ.0813923ÿÿÿÿÿ1.89ÿÿÿ0.079ÿÿÿÿ-.0198802ÿÿÿÿ.3270869
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿÿ_consÿ|ÿÿÿÿÿ5.2499ÿÿÿ.1615507ÿÿÿÿ32.50ÿÿÿ0.000ÿÿÿÿÿ4.905563ÿÿÿÿ5.594238
-------------------------------------------------------------------------------

.ÿpredictÿdoubleÿxb,ÿxb

.ÿ
.ÿassertÿround(pred,ÿ0.00001)ÿ==ÿround(xb,ÿ0.00001)ÿifÿinlist(id,ÿ1,ÿ2)

.ÿ
.ÿexit

endÿofÿdo-file


.

[Goh jedrek]

Something like the following should work.

.ÿ
.ÿversionÿ17.0

.ÿ
.ÿclearÿ*

.ÿ
.ÿquietlyÿinputÿbyteÿidÿfloatÿtimeÿintÿconcÿstr8ÿdietÿ///
>ÿÿÿÿÿÿÿÿÿfloat(elimÿpredÿlnalcoholÿpred_rep)

.ÿ
.ÿlabelÿdefineÿGuarStatusÿ0ÿnoGuarÿ1ÿwithGuar

.ÿencodeÿdiet,ÿgenerate(dit)ÿlabel(GuarStatus)ÿnoextend

.ÿ
.ÿregressÿlnalcoholÿc.time##i.dit##i.idÿifÿinlist(id,ÿ1,ÿ2)

ÿÿÿÿÿÿSourceÿ|ÿÿÿÿÿÿÿSSÿÿÿÿÿÿÿÿÿÿÿdfÿÿÿÿÿÿÿMSÿÿÿÿÿÿNumberÿofÿobsÿÿÿ=ÿÿÿÿÿÿÿÿ23
-------------+----------------------------------ÿÿÿF(7,ÿ15)ÿÿÿÿÿÿÿÿ=ÿÿÿÿÿ94.19
ÿÿÿÿÿÿÿModelÿ|ÿÿ2.60723948ÿÿÿÿÿÿÿÿÿ7ÿÿ.372462783ÿÿÿProbÿ>ÿFÿÿÿÿÿÿÿÿ=ÿÿÿÿ0.0000
ÿÿÿÿResidualÿ|ÿÿÿ.05931505ÿÿÿÿÿÿÿÿ15ÿÿ.003954337ÿÿÿR-squaredÿÿÿÿÿÿÿ=ÿÿÿÿ0.9778
-------------+----------------------------------ÿÿÿAdjÿR-squaredÿÿÿ=ÿÿÿÿ0.9674
ÿÿÿÿÿÿÿTotalÿ|ÿÿ2.66655453ÿÿÿÿÿÿÿÿ22ÿÿ.121207024ÿÿÿRootÿMSEÿÿÿÿÿÿÿÿ=ÿÿÿÿ.06288

-------------------------------------------------------------------------------
ÿÿÿÿlnalcoholÿ|ÿCoefficientÿÿStd.ÿerr.ÿÿÿÿÿÿtÿÿÿÿP>|t|ÿÿÿÿÿ[95%ÿconf.ÿinterval]
--------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿtimeÿ|ÿÿ-.3131906ÿÿÿÿ.039771ÿÿÿÿ-7.87ÿÿÿ0.000ÿÿÿÿ-.3979605ÿÿÿ-.2284207
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿÿÿÿditÿ|
ÿÿÿÿwithGuarÿÿ|ÿÿÿ-.058114ÿÿÿ.2902195ÿÿÿÿ-0.20ÿÿÿ0.844ÿÿÿÿ-.6767022ÿÿÿÿ.5604743
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿdit#c.timeÿ|
ÿÿÿÿwithGuarÿÿ|ÿÿÿ.0249545ÿÿÿ.0688854ÿÿÿÿÿ0.36ÿÿÿ0.722ÿÿÿÿ-.1218714ÿÿÿÿ.1717803
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿÿÿ2.idÿ|ÿÿÿ.3710292ÿÿÿ.2284672ÿÿÿÿÿ1.62ÿÿÿ0.125ÿÿÿÿÿ-.115937ÿÿÿÿ.8579955
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿid#c.timeÿ|
ÿÿÿÿÿÿÿÿÿÿÿ2ÿÿ|ÿÿ-.0876989ÿÿÿ.0562447ÿÿÿÿ-1.56ÿÿÿ0.140ÿÿÿÿ-.2075817ÿÿÿÿ.0321839
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿdit#idÿ|
ÿÿwithGuar#2ÿÿ|ÿÿÿ-.420878ÿÿÿÿ.336965ÿÿÿÿ-1.25ÿÿÿ0.231ÿÿÿÿ-1.139102ÿÿÿÿÿ.297346
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
dit#id#c.timeÿ|
ÿÿwithGuar#2ÿÿ|ÿÿÿ.1536034ÿÿÿ.0813923ÿÿÿÿÿ1.89ÿÿÿ0.079ÿÿÿÿ-.0198802ÿÿÿÿ.3270869
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿ|
ÿÿÿÿÿÿÿÿ_consÿ|ÿÿÿÿÿ5.2499ÿÿÿ.1615507ÿÿÿÿ32.50ÿÿÿ0.000ÿÿÿÿÿ4.905563ÿÿÿÿ5.594238
-------------------------------------------------------------------------------

.ÿpredictÿdoubleÿxb,ÿxb

.ÿ
.ÿassertÿround(pred,ÿ0.00001)ÿ==ÿround(xb,ÿ0.00001)ÿifÿinlist(id,ÿ1,ÿ2)

.ÿ
.ÿexit

endÿofÿdo-file


.

I have tried this and it doesn't work.

My issue is with grouping the variables together for the predict command to give the correct output.

I have tried doing this and it still doesn't work

gen participant_dietgroup = .
replace participant_dietgroup = 1 if diet == "withGuar" & id == 1 
replace participant_dietgroup = 2 if diet == "noGuar" & id == 1
replace participant_dietgroup = 3 if diet == "withGuar" & id == 2
replace participant_dietgroup = 4 if diet == "noGuar" & id == 2
replace participant_dietgroup = 5 if diet == "withGuar" & id == 3 
replace participant_dietgroup = 6 if diet == "noGuar" & id == 3
replace participant_dietgroup = 7 if diet == "withGuar" & id == 4
replace participant_dietgroup = 8 if diet == "noGuar" & id == 4

regress lnalcohol time if participant_dietgroup == 1 & participant_dietgroup == 2

I get the error 2000 and if I regress for just the first group, then I only get accurate output for the first group, while the other groups are inaccurate.

[Joseph Coveney]

I have tried this and it doesn't work.

Um, what I showed is output, ergo, I just showed you that it does work.

My issue is with grouping the variables together for the predict command to give the correct output.

I have tried doing this and it still doesn't work[code redacted]

You don't need to do that. In fact, as you discovered, if you try to do that, then it will thwart your intention.

Instead, form the proper interaction terms and fit a single model, just as I showed above. Then there need be only a single -predict- postestimation command, and it will match your earlier predict variable, just as I showed it does in your example.