Legend explanation does not show the same colors as in graph

[Anna Stubbendorff]

Hi!

Why do the legend explanations not have the same colors as in the graph?

I am using this code:

/// Loop through your variables
foreach y in Meanthickness72 {

// First regression with 'E'
regress `y' E i.sex2
local betahat1 = `:di %5.3f _b[E]'
matrix T = r(table) // Store regression table in matrix T
local p_value1 = `:di %05.3f T[1,3]' // Extract p-value for 'E' from the matrix

// Store the results for later use with coefplot
quietly eststo E: margins, at(E=(0(1)10)) post

// regression with 'M'
regress `y' M i.sex2
local betahat2 = `:di %5.3f _b[M]'
matrix T = r(table) // Store regression table in matrix T
local p_value2 = `:di %05.3f T[1,3]' // Extract p-value for 'M' from the matrix

// Store the results for later use with coefplot
quietly eststo M: margins, at(M=(0(1)10)) post

// Coefplot command with updated legend
coefplot E M, ///
at ///
title("`: var label `y''") ///
ytitle("`: var label `y''") ///
xtitle(Score) ///

legend(order( ///
1 "EAT-Lancet (B{sub:1}=`betahat1', p=`p_value1')" ///
2 "Mediterranean (B{sub:2}=`betahat2', p=`p_value2')" ///
) pos(6) col(2)) ///
recast(line)
}

coefplot (M, recast(connected) lc(red)) (E, recast(connected) lc(blue)), drop(_cons)


foreach y in Meanthickness72 {

// First regression with 'M'
regress `y' M i.sex2
local betahat1 = `:di %5.3f _b[M]'
matrix T = r(table) // Store regression table in matrix T
local p_value1 = `:di %05.3f T[1,3]' // Extract p-value for 'M' from the matrix

// Store the results for later use with coefplot
quietly eststo M: margins, at(M=(0(1)10)) post

// Second regression with 'E'
regress `y' E i.sex2
local betahat2 = `:di %5.3f _b[E]'
matrix T = r(table) // Store regression table in matrix T
local p_value2 = `:di %05.3f T[1,3]' // Extract p-value for 'E' from the matrix

// Store the results for later use with coefplot
quietly eststo E: margins, at(E=(0(1)10)) post

coefplot M E, ///
at ///
title("`: var label `y''") ///
ytitle("`: var label `y''") ///
xtitle(Score) ///
recast(line) ///
mstyle(M, color(blue) lpattern(solid)) /// Specify color and line pattern for M
mstyle(E, color(red) lpattern(dash)) /// Specify color and line pattern for E
legend(order( ///
1 "Mediterranean (B{sub:1}=`betahat1', p=`p_value1')" ///
2 "EAT-Lancet (B{sub:2}=`betahat2', p=`p_value2')" ///
) pos(6) col(2))
}

[Scott Merryman]

Your graph is made up of 4 plots : 2 -rcap- for the confidence intervals (which are plots 1 & 2) and 2 -line-. Try:

legend(order(3 "Medi ... "  4 "EAT ..." )

[Anna Stubbendorff]

Thank you!!

I updated the code with 2 and 3, which made it work. But is there a way to simplify this code?

/// Loop through your variables for regression and graphs.
foreach y in Meanthickness72 GCS1 AD_MRIsignature_Jack_adjusted Hippadj72 gap_ensemble {

// First regression with 'Eat-Lancet diet'
regress `y' E i.sex2 Rök_dic_2 Physact2 BMI3 EDU3
local betahat1 = `:di %5.3f _b[E]'
matrix T = r(table) // Store regression table in matrix T
local p_value1 = `:di %05.3f T[4,1]' // Extract p-value for 'E' from the matrix
// Store the results for later use with coefplot
quietly eststo E: margins, at(E=(0(1)10)) post

// regression with 'Mediterranean diet'
regress `y' M i.sex2 Rök_dic_2 Physact2 BMI3 EDU3
local betahat2 = `:di %5.3f _b[M]'
matrix T = r(table) // Store regression table in matrix T
local p_value2 = `:di %05.3f T[4,1]' // Extract p-value for 'M' from the matrix
// Store the results for later use with coefplot
quietly eststo M: margins, at(M=(0(1)10)) post

// Coefplot command with updated legend
coefplot E M, ///
at ///
title("`: var label `y''") ///
ytitle("`: var label `y''") ///
xtitle(Score) ///
legend(order( ///
2 "EAT-Lancet (B{sub:1}=`betahat1', p=`p_value1')" ///
3 "Mediterranean (B{sub:2}=`betahat2', p=`p_value2')" ///
) pos(6) col(2)) recast(line)
graph save `y'.gph, replace
graph export `y'.jpg, quality (100) replace
}

[Nick Cox]

I recommend statements like

local betahat1 : di %5.3f _b[E]

rather than

local betahat1 = `:di %5.3f _b[E]'

That's only a saving of 3 characters, but three characters you can get wrong too.