How to statistically compare two curvilinear lines in a single plot using DO LOOP

[SASNovice2014]

Hello,

I have an unbalanced panel dataset across 10 years as follows. Here are my variables:

Y= dependent variable

x1= predictor

x1_2 = predictor (squared)

x3 = moderator

name	time	gvkey	y	x1	x1_2	x3
abc	1	10553	0.95	-0.20	0.04	0.34
abc	2	10553	1.20
abc	3	10553	1.29	-0.45	0.20	0.44
abc	4	10553	1.62	-0.32	0.10	0.23
…
xyz	1	7435	1.97	-1.09	1.19
xyz	2	7435	1.86	-0.69	0.48	0.41
xyz	3	7435	1.90	-0.70	0.49	0.42
xyz	4	7435	1.94	-1.90	3.62	2.67

Here's SAS codes to run regression where x3 has been split into high sample:

proc panel data=test_xlag ;

inst depvar

exogenous=( x3_high )

pred=(x1  x1_2)  ;

model y_high =  x3_high   x1   x1_2 

/ gmm  twostep maxband=1 time robust ;

     id gvkey time;

run;

Here's SAS codes to run regression where x3 has been split into low sample:

proc panel data=test_xlag ;

inst depvar

exogenous=( x3_low )

pred=(x1  x1_2)  ;

model y_low = x3_low   x1   x1_2 

/ gmm  twostep maxband=1 time robust ;

     id gvkey time;

run;

Here's SAS codes to run both regressions in a DO LOOP:

data test_xlag;

do X1 = -3.0 to 0 by 0.1;

      y_high = 0.136544 + (-0.04991)*X1 + (-0.00773)*X1*X1 ;

      y_low  = 0.152363 + (-0.05342)*X1 + (-0.00836)*X1*X1 ;

    output;

    end;

run;

Let's assume I've created the ANNOTATE dataset and produce the plot in the attached file. The plot is an inverse U-shaped curve. It has 2 parallel curvilinear lines (red curve = depicts regression based on high sample, blue curve = depicts regression based on low sample). In the plot, x axis is range of x1 values, and y axis is range of y values.

Here are my questions:

(1) How do create a DO LOOP in SAS 9.3 that would allow me to statistically compare the red line to the blue line for x1 ranging from -0.1 to -1.0 at increments of 0.1? How do I test whether the red line is statistically different from the blue line at each specified values of x1? I would need chi-square values and p-values for each set of red/blue comparisons at each of the ten x1 values.

(2) How do I generate marginal effects and standard errors for each of blue line and red line at each of the 10 specified x1 values (for a total of 20 values)? What SAS procedure do I use?

Below are the STATA codes to generate what I need for (1) and (2). Can someone help me translate these STATA codes into SAS codes?

margins, at(x1 = (-0.1 (0.1) -1.0) x3_high=(0 1 )) post

margins, coeflegend

forvalues i = 1(2)20 {

loc j = `i' +1

test _b[`i'._at] = _b[`j'._at]

}

Thanks in advance!

[SASNovice2014]

Hello,

I have an unbalanced panel dataset across 10 years as follows. Here are my variables:

Y= dependent variable

x1= predictor

x1_2 = predictor (squared)

x3 = moderator

name	time	gvkey	y	x1	x1_2	x3
abc	1	10553	0.95	-0.20	0.04	0.34
abc	2	10553	1.20
abc	3	10553	1.29	-0.45	0.20	0.44
abc	4	10553	1.62	-0.32	0.10	0.23
…
xyz	1	7435	1.97	-1.09	1.19
xyz	2	7435	1.86	-0.69	0.48	0.41
xyz	3	7435	1.90	-0.70	0.49	0.42
xyz	4	7435	1.94	-1.90	3.62	2.67

Here's SAS codes to run regression where x3 has been split into high sample:

proc panel data=test_xlag ;

inst depvar

exogenous=( x3_high )

pred=(x1  x1_2)  ;

model y_high =  x3_high   x1   x1_2 

/ gmm  twostep maxband=1 time robust ;

     id gvkey time;

run;

Here's SAS codes to run regression where x3 has been split into low sample:

proc panel data=test_xlag ;

inst depvar

exogenous=( x3_low )

pred=(x1  x1_2)  ;

model y_low = x3_low   x1   x1_2 

/ gmm  twostep maxband=1 time robust ;

     id gvkey time;

run;

Here's SAS codes to run both regressions in a DO LOOP:

data test_xlag;

do X1 = -3.0 to 0 by 0.1;

      y_high = 0.136544 + (-0.04991)*X1 + (-0.00773)*X1*X1 ;

      y_low  = 0.152363 + (-0.05342)*X1 + (-0.00836)*X1*X1 ;

    output;

    end;

run;

Let's assume I've created the ANNOTATE dataset and produce the plot in the attached file. The plot is an inverse U-shaped curve. It has 2 parallel curvilinear lines (red curve = depicts regression based on high sample, blue curve = depicts regression based on low sample). In the plot, x axis is range of x1 values, and y axis is range of y values.

Here are my questions:

(1) How do create a DO LOOP in SAS 9.3 that would allow me to statistically compare the red line to the blue line for x1 ranging from -0.1 to -1.0 at increments of 0.1? How do I test whether the red line is statistically different from the blue line at each specified values of x1? I would need chi-square values and p-values for each set of red/blue comparisons at each of the ten x1 values.

(2) How do I generate marginal effects and standard errors for each of blue line and red line at each of the 10 specified x1 values (for a total of 20 values)? What SAS procedure do I use?

Below are the STATA codes to generate what I need for (1) and (2). Can someone help me translate these STATA codes into SAS codes?

margins, at(x1 = (-0.1 (0.1) -1.0) x3_high=(0 1 )) post

margins, coeflegend

forvalues i = 1(2)20 {

loc j = `i' +1

test _b[`i'._at] = _b[`j'._at]

}

Thanks in advance!

[SASNovice2014]

Hello,

I have an unbalanced panel dataset across 10 years as follows. Here are my variables:

Y= dependent variable

x1= predictor

x1_2 = predictor (squared)

x3 = moderator

name	time	gvkey	y	x1	x1_2	x3
abc	1	10553	0.95	-0.20	0.04	0.34
abc	2	10553	1.20
abc	3	10553	1.29	-0.45	0.20	0.44
abc	4	10553	1.62	-0.32	0.10	0.23
…
xyz	1	7435	1.97	-1.09	1.19
xyz	2	7435	1.86	-0.69	0.48	0.41
xyz	3	7435	1.90	-0.70	0.49	0.42
xyz	4	7435	1.94	-1.90	3.62	2.67

Here's SAS codes to run regression where x3 has been split into high sample:

proc panel data=test_xlag ;

inst depvar

exogenous=( x3_high )

pred=(x1  x1_2)  ;

model y_high =  x3_high   x1   x1_2 

/ gmm  twostep maxband=1 time robust ;

     id gvkey time;

run;

Here's SAS codes to run regression where x3 has been split into low sample:

proc panel data=test_xlag ;

inst depvar

exogenous=( x3_low )

pred=(x1  x1_2)  ;

model y_low = x3_low   x1   x1_2 

/ gmm  twostep maxband=1 time robust ;

     id gvkey time;

run;

Here's SAS codes to run both regressions in a DO LOOP:

data test_xlag;

do X1 = -3.0 to 0 by 0.1;

      y_high = 0.136544 + (-0.04991)*X1 + (-0.00773)*X1*X1 ;

      y_low  = 0.152363 + (-0.05342)*X1 + (-0.00836)*X1*X1 ;

    output;

    end;

run;

Let's assume I've created the ANNOTATE dataset and produce the plot in the attached file. The plot is an inverse U-shaped curve. It has 2 parallel curvilinear lines (red curve = depicts regression based on high sample, blue curve = depicts regression based on low sample). In the plot, x axis is range of x1 values, and y axis is range of y values.

Here are my questions:

(1) How do create a DO LOOP in SAS 9.3 that would allow me to statistically compare the red line to the blue line for x1 ranging from -0.1 to -1.0 at increments of 0.1? How do I test whether the red line is statistically different from the blue line at each specified values of x1? I would need chi-square values and p-values for each set of red/blue comparisons at each of the ten x1 values.

(2) How do I generate marginal effects and standard errors for each of blue line and red line at each of the 10 specified x1 values (for a total of 20 values)? What SAS procedure do I use?

Below are the STATA codes to generate what I need for (1) and (2). Can someone help me translate these STATA codes into SAS codes?

margins, at(x1 = (-0.1 (0.1) -1.0) x3_high=(0 1 )) post

margins, coeflegend

forvalues i = 1(2)20 {

loc j = `i' +1

test _b[`i'._at] = _b[`j'._at]

}

Thanks in advance!

[SASNovice2014]

Hello,

I have an unbalanced panel dataset across 10 years as follows. Here are my variables:

Y= dependent variable

x1= predictor

x1_2 = predictor (squared)

x3 = moderator

name	time	gvkey	y	x1	x1_2	x3
abc	1	10553	0.95	-0.20	0.04	0.34
abc	2	10553	1.20
abc	3	10553	1.29	-0.45	0.20	0.44
abc	4	10553	1.62	-0.32	0.10	0.23
…
xyz	1	7435	1.97	-1.09	1.19
xyz	2	7435	1.86	-0.69	0.48	0.41
xyz	3	7435	1.90	-0.70	0.49	0.42
xyz	4	7435	1.94	-1.90	3.62	2.67

Here's SAS codes to run regression where x3 has been split into high sample:

proc panel data=test_xlag ;

inst depvar

exogenous=( x3_high )

pred=(x1  x1_2)  ;

model y_high =  x3_high   x1   x1_2 

/ gmm  twostep maxband=1 time robust ;

     id gvkey time;

run;

Here's SAS codes to run regression where x3 has been split into low sample:

proc panel data=test_xlag ;

inst depvar

exogenous=( x3_low )

pred=(x1  x1_2)  ;

model y_low = x3_low   x1   x1_2 

/ gmm  twostep maxband=1 time robust ;

     id gvkey time;

run;

Here's SAS codes to run both regressions in a DO LOOP:

data test_xlag;

do X1 = -3.0 to 0 by 0.1;

      y_high = 0.136544 + (-0.04991)*X1 + (-0.00773)*X1*X1 ;

      y_low  = 0.152363 + (-0.05342)*X1 + (-0.00836)*X1*X1 ;

    output;

    end;

run;

Let's assume I've created the ANNOTATE dataset and produce the plot in the attached file. The plot is an inverse U-shaped curve. It has 2 parallel curvilinear lines (red curve = depicts regression based on high sample, blue curve = depicts regression based on low sample). In the plot, x axis is range of x1 values, and y axis is range of y values.

Here are my questions:

(1) How do create a DO LOOP in SAS 9.3 that would allow me to statistically compare the red line to the blue line for x1 ranging from -0.1 to -1.0 at increments of 0.1? How do I test whether the red line is statistically different from the blue line at each specified values of x1? I would need chi-square values and p-values for each set of red/blue comparisons at each of the ten x1 values.

(2) How do I generate marginal effects and standard errors for each of blue line and red line at each of the 10 specified x1 values (for a total of 20 values)? What SAS procedure do I use?

Below are the STATA codes to generate what I need for (1) and (2). Can someone help me translate these STATA codes into SAS codes?

margins, at(x1 = (-0.1 (0.1) -1.0) x3_high=(0 1 )) post

margins, coeflegend

forvalues i = 1(2)20 {

loc j = `i' +1

test _b[`i'._at] = _b[`j'._at]

}

Thanks in advance!