[R] Running segmented on grouped data and collating model parameters in a data frame

[Sumitrajit Dhar]

Hi Folks,

I am trying to teach myself how to solve the problem described below but am
running out of time. Hence the plea for help. Thanks in advance.

Here is my data frame.

> t
# A tibble: 12 x 12
   subject  ageGrp ear   hearingGrp sex    freq    L2   Ldp Phidp    NF
SNR   Ldp_p
   <fct>    <fct>  <fct> <fct>      <fct> <int> <int> <dbl> <dbl> <dbl>
<dbl>   <dbl>
 1 HALAF573 A      L     A          F         2     0 -19.6 197.  -28.5
8.88  2.10e-6
 2 HALAF573 A      L     A          F         2     2 -18.7 203.  -22.0
3.25  2.31e-6
 3 HALAF573 A      L     A          F         2     4 -29.1 255.  -27.4
 -1.64  7.04e-7
 4 HALAF573 A      L     A          F         2     6 -12.4 174.  -12.2
 -0.206 4.78e-6
 5 HALAF573 A      L     A          F         4     0 -28.6 232.  -26.7
 -1.87  7.45e-7
 6 HALAF573 A      L     A          F         4     2 -27.2 351.  -28.8
1.59  8.68e-7
 7 HALAF573 A      L     A          F         4     4 -20.4  26.2 -35.0
 14.6   1.92e-6
 8 HALAF573 A      L     A          F         4     6 -20.0  85.1 -29.8
9.75  2.00e-6
 9 HALAF573 A      L     A          F         8     0 -22.8  39.2 -22.1
 -0.689 1.45e-6
10 HALAF573 A      L     A          F         8     2 -14.5  13.4 -20.7
6.23  3.76e-6
11 HALAF573 A      L     A          F         8     4 -17.3 345.  -21.6
4.30  2.73e-6
12 HALAF573 A      L     A          F         8     6 -14.1 320.  -21.7
7.59  3.94e-6

# Note that there are more levels of L2 (31 in total)  and 344 other
subjects but I truncated the frame for posting.

# I want to do this:  t %>%  group_by(freq) %>% [run segmented] %>% [create
a data frame with [subject, freq, breakpoint1, breakpoint2, slope1, slope2,
slope3, L2 when Ldp_p == 0].

# Also note that ultimately I will be grouping by "subject, freq".

# I can run the models and get believable results. The following run on a
data frame with L2 between 0 and 60.

out.lm <- lm(Ldp_p ~ L2, data = t)
o <- segmented(out.lm, seg.Z = ~L2, psi = list(L2 = c(20,45)),
               control = seg.control(display = FALSE)

o$psi
#        Initial     Est.    St.Err
#psi1.L2      20 30.78256 0.5085192
#psi2.L2      45 53.16390 0.4671701

slope(o)
slope(o)
$L2
#              Est.    St.Err. t value   CI(95%).l   CI(95%).u
#slope1  1.2060e-06 1.6606e-07  7.2622  8.6397e-07  1.5480e-06
#slope2  1.0708e-05 2.9196e-07 36.6770  1.0107e-05  1.1309e-05
#slope3 -4.5791e-06 1.3694e-06 -3.3439 -7.3995e-06 -1.7588e-06

Thanks again for bailing me out.

Regards,
Sumit

	[[alternative HTML version deleted]]