#!/usr/bin/env bash
# Testing gmtregress on the data in Draper & Smith [1998]
# Applied Regression Analysis, 3rd Ed, Wiley.

ps=draper.ps
cat << EOF > draper.txt
# Table 1.1 in Draper & Smith, Applied Regression Analysis
35.3	10.98
29.7	11.13
30.8	12.51
58.8	8.40
61.4	9.27
71.3	8.73
74.4	6.36
76.7	8.50
70.7	7.82
57.5	9.14
46.4	8.24
28.9	12.19
28.1	11.88
39.1	9.57
46.8	10.94
48.5	9.58
59.3	10.09
70.0	8.11
70.0	6.83
74.5	8.88
72.1	7.68
58.1	8.47
44.6	8.86
33.4	10.36
28.6	11.08
EOF
# First plot data and basic LS fit with equation
txt=$(gmt regress -Ey -N2 -Fxm draper.txt -T0 | awk '{printf "85 13 @!y\\303 = %.4f %.4f x\n", $17, $15}')
gmt psbasemap -R20/85/6/13 -JX6.5i/4i -P -Xc -K -Baf -BWSne > $ps
gmt regress -Ey -N2 -Fxym draper.txt | awk '{printf "> error\n%s %s\n%s %s\n", $1, $2, $1, $3}' | gmt psxy -R -J -O -K -W0.25p,red,- >> $ps
gmt psxy -R -J -O -K draper.txt -Sc0.2c -Gblue >> $ps
gmt regress -Ey -N2 -Fxm -T25/80/2+n draper.txt | gmt psxy -R -J -O -K -W2p >> $ps
echo "$txt" | gmt pstext -R -J -O -K -F+jRT+f18p -Dj0.1i >> $ps
# Redo plot and basic LS fit but also show 68%, 95% & 99% confidence band
gmt psbasemap -R -J -O -K -Baf -BWSNe+t"Draper & Smith [1998] Regression" -Y4.75i >> $ps
gmt regress -Ey -N2 -Fxmc -T25/80/1 -C99 draper.txt | gmt psxy -R -J -O -K -L+d -Glightgreen >> $ps
gmt regress -Ey -N2 -Fxmc -T25/80/1 -C95 draper.txt | gmt psxy -R -J -O -K -L+d -Glightorange >> $ps
gmt regress -Ey -N2 -Fxmc -T25/80/1 -C68 draper.txt | gmt psxy -R -J -O -K -L+d -Glightred -W2p >> $ps
gmt psxy -R -J -O -K draper.txt -Sc0.2c -Gblue >> $ps
gmt pslegend -DjTR+w1.65i+jRT+o0.1i/0.1i -R -J -O -F+p1p << EOF >> $ps
S 0.1i s 0.15i lightgreen  - 0.25i 99% Confidence
S 0.1i s 0.15i lightorange - 0.25i 95% Confidence
S 0.1i s 0.15i lightred    - 0.25i 68% Confidence
EOF