[afro-nets] Assistance on Logistic Regression

[Daniel Chirundu <dchirundu@comone.co.zw>]

Assistance on Logistic Regression
---------------------------------

I was doing an investigation of the risk factors for schisto-
somiasis (S. haematobium).

I found significant associations at 95% confidences between hav-
ing a current S. haematobium infection as diagnosed by the sy-
ringe filtration method and

Using a net to capture fish (N29)
Ability to fish using a fishing line (N26)
Ability to swim (N22)
Participation in watering the household garden (N8)

I had found that

Sex (N1) and Doing Laundry at the river (N31) were effect modi-
fier (interactions)

I put this in a logistic regression model and obtained the out-
put below:

Epi Info

Results Library
Current View:
C:\Epi_Info\Field Studies\Shurugwi Schisto Sur-
vey.MDB:viewSchisto
Record Count:
277
(Deleted records excluded)

Date:
8/3/2005 4:18:20 PM
LOGISTIC SHaematobium = N1Sex N22CanYouSwim N28Fishing
N29DoYouUse N31InLastMonth N8DoYouPartcipate
N1Sex*N31InLastMonth

Unconditional Logistic Regression

Term
Odds Ratio
95%
C.I.
Coefficient
S. E.
Z-Statistic
P-Value
N1Sex (Male/Female)
4.5742
0.5566
37.5905
1.5204
1.0747
1.4148
0.1571
N22CanYouSwim (Yes/No)
1.0103
0.4492
2.2726
0.0103
0.4136
0.0248
0.9802
N28Fishing (Yes/No)
0.8629
0.4039
1.8436
-0.1474
0.3873
-0.3806
0.7035
N29DoYouUse (Yes/No)
1.9914
0.8448
4.6941
0.6888
0.4375
1.5744
0.1154
N31InLastMonth (Yes/No)
2.3840
0.3279
17.3336
0.8688
1.0122
0.8583
0.3907
N8DoYouPartcipate (Yes/No)
6.6314
1.1471
38.3351
1.8918
0.8952
2.1133
0.0346
N1Sex (Male/Female) * N31InLastMonth (Yes/No)
0.3646
0.0398
3.3424
-1.0090
1.1304
-0.8925
0.3721
CONSTANT
*
*
*
-2.1381
1.2529
-1.7064
0.0879


Convergence:
Converged
Iterations:
5
Final -2*Log-Likelihood:
207.3426
Cases included:
181

Test
Statistic
D.F.
P-Value
Score
11.9675
7
0.1016
Likelihood Ratio
11.5426
7
0.1166

I then removed the non significant variables till I got the out-
put below:

Previous Dataset Results Library
LOGISTIC SHaematobium = N1Sex N29DoYouUse N31InLastMonth
N8DoYouPartcipate N1Sex*N31InLastMonth

Next Procedure

Unconditional Logistic Regression

Term
Odds Ratio
95%
C.I.
Coefficient
S. E.
Z-Statistic
P-Value
N1Sex (Male/Female)
4.5120
1.0365
19.6408
1.5068
0.7505
2.0078
0.0447
N29DoYouUse (Yes/No)
2.3347
1.1221
4.8580
0.8479
0.3738
2.2680
0.0233
N31InLastMonth (Yes/No)
1.8227
0.5023
6.6144
0.6003
0.6576
0.9129
0.3613
N8DoYouPartcipate (Yes/No)
9.4815
1.7763
50.6107
2.2493
0.8545
2.6323
0.0085
N1Sex (Male/Female) * N31InLastMonth (Yes/No)
0.3840
0.0778
1.8960
-0.9571
0.8147
-1.1747
0.2401
CONSTANT
*
*
*
-2.4778
1.0258
-2.4154
0.0157


Convergence:
Converged
Iterations:
5
Final -2*Log-Likelihood:
287.7458
Cases included:
245

Test
Statistic
D.F.
P-Value
Score
20.1113
5
0.0012
Likelihood Ratio
20.3180
5
0.0011

My interpretation of this is:

In a logistic regression model, taking sex and doing laundry at
the river as effect modifiers, being male (N1), using nets to
capture fish (n29), participation in watering the household gar-
den (N8) remained statistically significant at 95% confidence.

1. Is this the correct way to do it?
2. And is the interpretation correct?

--
Daniel Chirundu
mailto:dchirundu@comone.co.zw