# Help with testing three variables (non-metrical) perhaps GEE

#### rokochan95

##### New Member
Hi,
I ran a survey (n=38) and now I'm trying to analyze one hypothesis with Rstudio:

H4: People workout more when working from home, than in office.

I'm trying to combine two items to test this hypothesis.
1. sporting activity (frequency per week; ordinal scale) in 4 different periods (pre-covid, first lockdown, between lockdowns, second lockdown)
2. homeoffice [homeoffice, no homeoffice] in 4 different periods (as above)

I tried to do a GEE-test (someone recommended that). For that I created another variable time [1;2;3;4] for the periods, but not sure if I'm on the right track.
Thing is, I don't know nothing about GEE, but it seems the only way.

1. The created the dataset:
Code:
id = subject [1,38], time = periods [1,4], sport = frequency per week [0,4], homeoffice [0,1]

# A tibble: 6 x 4
id  time sport homeoffice
<dbl> <dbl> <dbl> <fct>
1     1     1     2 1
2     2     1     4 0
3     3     1     4 0
4     4     1     2 0
5     5     1     3 0
6     6     1     2 0
2. Then I set a factor & formula (I don't know, if either are correct) and ran geeglm:
Code:
gee$homeoffice <- as.factor(gee$homeoffice)
mf <- formula(sport ~ time + homeoffice)
geeInd <- geeglm(mf, id=id, data=gee, family=gaussian, corstr="ind")
summary(geeInd)

##I don't know, if the "arguments" id, family, corstr are correctly chosen

Coefficients:
Estimate Std.err  Wald Pr(>|W|)
(Intercept)   2.2927  0.2574 79.34   <2e-16 ***
time         -0.1309  0.0911  2.06   0.1507
homeoffice1  -0.7092  0.2190 10.49   0.0012 **
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Correlation structure = independence
Estimated Scale Parameters:

Estimate Std.err
(Intercept)     1.41   0.142
Number of clusters:   152  Maximum cluster size: 1
I have no clue if anything in this procedure is correct or if GEE is even the right model I'm supposed to use.
I would appreciate any help. Looking forward.

Kind regards,
rokochan95