Feature #2924 sal1l2_mae, PR 3 of 3

docs/Users_Guide/appendixC.rst

@@ -616,23 +616,23 @@ Anomaly Correlation Coefficient
 
 Called "ANOM_CORR" and "ANOM_CORR_UNCNTR" for centered and uncentered versions in CNT output :numref:`table_PS_format_info_CNT`
 
-The anomaly correlation coefficient is equivalent to the Pearson correlation coefficient, except that both the forecasts and observations are first adjusted according to a climatology value. The anomaly is the difference between the individual forecast or observation and the typical situation, as measured by a climatology (**c**) of some variety. It measures the strength of linear association between the forecast anomalies and observed anomalies. The anomaly correlation coefficient is defined as: 
+The anomaly correlation coefficient is equivalent to the Pearson correlation coefficient, except that both the forecasts and observations are first adjusted by subtracting their corresponding climatology value. The anomaly is the difference between the individual forecast or observation and the typical situation, as measured by a forecast climatology (:math:`c_f`) and observation climatology (:math:`c_o`). It measures the strength of linear association between the forecast anomalies and observed anomalies. The anomaly correlation coefficient is defined as:
 
-.. math:: \text{Anomaly Correlation} = \frac{\sum(f_i - c)(o_i - c)}{\sqrt{\sum(f_i - c)^2} \sqrt{\sum(o_i -c)^2}} .
+.. math:: \text{Anomaly Correlation} = \frac{\sum(f_i - {c_f}_i)(o_i - {c_o}_i)}{\sqrt{\sum(f_i - {c_f}_i)^2} \sqrt{\sum(o_i - {c_o}_i)^2}} .
 
 The centered anomaly correlation coefficient (ANOM_CORR) which includes the mean error is defined as: 
 
 .. only:: latex
 
-  .. math:: \text{ANOM\_CORR } = \frac{ \overline{[(f - c) - \overline{(f - c)}][(a - c) - \overline{(a - c)}]}}{ \sqrt{ \overline{( (f - c) - \overline{(f - c)})^2} \overline{( (a - c) - \overline{(a - c)})^2}}}
+  .. math:: \text{ANOM\_CORR } = \frac{ \overline{[(f - c_f) - \overline{(f - c_f)}][(o - c_o) - \overline{(o - c_o)}]}}{ \sqrt{ \overline{( (f - c_f) - \overline{(f - c_f)})^2} \overline{( (o - c_o) - \overline{(o - c_o)})^2}}}
 
 .. only:: html
 
-  .. math:: \text{ANOM_CORR } = \frac{ \overline{[(f - c) - \overline{(f - c)}][(a - c) - \overline{(a - c)}]}}{ \sqrt{ \overline{( (f - c) - \overline{(f - c)})^2} \overline{( (a - c) - \overline{(a - c)})^2}}}
+  .. math:: \text{ANOM_CORR } = \frac{ \overline{[(f - c_f) - \overline{(f - c_f)}][(o - c_o) - \overline{(o - c_o)}]}}{ \sqrt{ \overline{( (f - c_f) - \overline{(f - c_f)})^2} \overline{( (o - c_o) - \overline{(o - c_o)})^2}}}
 
 The uncentered anomaly correlation coefficient (ANOM_CORR_UNCNTR) which does not include the mean errors is defined as: 
 
-.. math:: \text{Anomaly Correlation Raw } = \frac{ \overline{(f - c)(a - c)}}{ \sqrt{\overline{(f - c)^2} \overline{(a - c)^2}}}
+.. math:: \text{Anomaly Correlation Raw } = \frac{ \overline{(f - c_f)(o - c_o)}}{ \sqrt{\overline{(f - c_f)^2} \overline{(o - c_o)^2}}}
 
 Anomaly correlation can range between -1 and 1; a value of 1 indicates perfect correlation and a value of -1 indicates perfect negative correlation. A value of 0 indicates that the forecast and observed anomalies are not correlated.
 
@@ -650,56 +650,60 @@ The partial sums can be accumulated over individual cases to produce statistics
 Scalar L1 and L2 Values
 -----------------------
 
-Called "FBAR", "OBAR", "FOBAR", "FFBAR", and "OOBAR" in SL1L2 output :numref:`table_PS_format_info_SL1L2`
+Called "FBAR", "OBAR", "FOBAR", "FFBAR", "OOBAR", and "MAE" in SL1L2 output :numref:`table_PS_format_info_SL1L2`
 
 These statistics are simply the 1st and 2nd moments of the forecasts, observations and errors:
 
 .. math:: 
-  \text{FBAR} = \text{Mean}(f) = \bar{f} = \frac{1}{n} \sum_{i=1}^n f_i
+  \text{FBAR} = \text{Mean}(f) = \frac{1}{n} \sum_{i=1}^n f_i
 
-  \text{OBAR} = \text{Mean}(o) = \bar{o} = \frac{1}{n} \sum_{i=1}^n o_i
+  \text{OBAR} = \text{Mean}(o) = \frac{1}{n} \sum_{i=1}^n o_i
 
-  \text{FOBAR} = \text{Mean}(fo) = \bar{fo} = \frac{1}{n} \sum_{i=1}^n f_i o_i
+  \text{FOBAR} = \text{Mean}(fo) = \frac{1}{n} \sum_{i=1}^n f_i o_i
 
-  \text{FFBAR} = \text{Mean}(f^2) = \bar{f}^2 = \frac{1}{n} \sum_{i=1}^n f_i^2
+  \text{FFBAR} = \text{Mean}(f^2) = \frac{1}{n} \sum_{i=1}^n f_i^2
 
-  \text{OOBAR} = \text{Mean}(o^2) = \bar{o}^2 = \frac{1}{n} \sum_{i=1}^n o_i^2
+  \text{OOBAR} = \text{Mean}(o^2) = \frac{1}{n} \sum_{i=1}^n o_i^2
+
+  \text{MAE} = \text{Mean}(|f - o|) = \frac{1}{n} \sum_{i=1}^n |f_i - o_i|
 
 Some of the other statistics for continuous forecasts (e.g., RMSE) can be derived from these moments.
 
 Scalar Anomaly L1 and L2 Values
 -------------------------------
 
-Called "FABAR", "OABAR", "FOABAR", "FFABAR", "OOABAR" in SAL1L2 output :numref:`table_PS_format_info_SAL1L2`
+Called "FABAR", "OABAR", "FOABAR", "FFABAR", "OOABAR", and "MAE" in SAL1L2 output :numref:`table_PS_format_info_SAL1L2`
 
-Computation of these statistics requires a climatological value, c. These statistics are the 1st and 2nd moments of the scalar anomalies. The moments are defined as:
+Computation of these statistics requires climatological values, where :math:`c_f` is the forecast climatology value and :math:`c_o` is the observation climatology value. These statistics are the 1st and 2nd moments of the scalar anomalies. The moments are defined as:
 
 .. math:: 
-  \text{FABAR} = \text{Mean}(f - c) = \bar{f - c} = \frac{1}{n} \sum_{i=1}^n (f_i - c)
+  \text{FABAR} = \text{Mean}(f - c_f) = \frac{1}{n} \sum_{i=1}^n (f_i - {c_f}_i)
+
+  \text{OABAR} = \text{Mean}(o - c_o) = \frac{1}{n} \sum_{i=1}^n (o_i - {c_o}_i)
 
-  \text{OABAR} = \text{Mean}(o - c) = \bar{o - c} = \frac{1}{n} \sum_{i=1}^n (o_i - c)
+  \text{FOABAR} = \text{Mean}[(f - c_f)(o - c_o)] = \frac{1}{n} \sum_{i=1}^n (f_i - {c_f}_i)(o_i - {c_o}_i)
 
-  \text{FOABAR} = \text{Mean}[(f - c)(o - c)] = \bar{(f - c)(o - c)} = \frac{1}{n} \sum_{i=1}^n (f_i - c)(o_i - c)
+  \text{FFABAR} = \text{Mean}[(f - c_f)^2] = \frac{1}{n} \sum_{i=1}^n (f_i - {c_f}_i)^2
 
-  \text{FFABAR} = \text{Mean}[(f - c)^2] = \bar{(f - c)}^2 = \frac{1}{n} \sum_{i=1}^n (f_i - c)^2
+  \text{OOABAR} = \text{Mean}[(o - c_o)^2] = \frac{1}{n} \sum_{i=1}^n (o_i - {c_o}_i)^2
 
-  \text{OOABAR} = \text{Mean}[(o - c)^2] = \bar{(o - c)}^2 = \frac{1}{n} \sum_{i=1}^n (o_i - c)^2
+  \text{MAE} = \text{Mean}(|(f - c_f) - (o - c_o)|) = \frac{1}{n} \sum_{i=1}^n |(f_i - {c_f}_i) - (o_i - {c_o}_i)|
 
 Vector L1 and L2 Values
 -----------------------
 
-Called "UFBAR", "VFBAR", "UOBAR", "VOBAR", "UVFOBAR", "UVFFBAR", "UVOOBAR" in VL1L2 output :numref:`table_PS_format_info_VL1L2`
+Called "UFBAR", "VFBAR", "UOBAR", "VOBAR", "UVFOBAR", "UVFFBAR", and "UVOOBAR" in VL1L2 output :numref:`table_PS_format_info_VL1L2`
 
-These statistics are the moments for wind vector values, where **u** is the E-W wind component and **v** is the N-S wind component ( :math:`u_f` is the forecast E-W wind component; :math:`u_o` is the observed E-W wind component; :math:`v_f` is the forecast N-S wind component; and :math:`v_o` is the observed N-S wind component). The following measures are computed:
+These statistics are the moments for wind vector values, where :math:`u` is the E-W wind component and :math:`v` is the N-S wind component ( :math:`u_f` is the forecast E-W wind component; :math:`u_o` is the observed E-W wind component; :math:`v_f` is the forecast N-S wind component; and :math:`v_o` is the observed N-S wind component). The following measures are computed:
 
 .. math:: 
-  \text{UFBAR} = \text{Mean}(u_f) = \bar{u}_f = \frac{1}{n} \sum_{i=1}^n u_{fi}
+  \text{UFBAR} = \text{Mean}(u_f) = \frac{1}{n} \sum_{i=1}^n u_{fi}
 
-  \text{VFBAR} = \text{Mean}(v_f) = \bar{v}_f = \frac{1}{n} \sum_{i=1}^n v_{fi}
+  \text{VFBAR} = \text{Mean}(v_f) = \frac{1}{n} \sum_{i=1}^n v_{fi}
 
-  \text{UOBAR} = \text{Mean}(u_o) = \bar{u}_o = \frac{1}{n} \sum_{i=1}^n u_{oi}
+  \text{UOBAR} = \text{Mean}(u_o) = \frac{1}{n} \sum_{i=1}^n u_{oi}
 
-  \text{VOBAR} = \text{Mean}(v_o) = \bar{v}_o = \frac{1}{n} \sum_{i=1}^n v_{oi}
+  \text{VOBAR} = \text{Mean}(v_o) = \frac{1}{n} \sum_{i=1}^n v_{oi}
 
   \text{UVFOBAR} = \text{Mean}(u_f u_o + v_f v_o) = \frac{1}{n} \sum_{i=1}^n (u_{fi} u_{oi} + v_{fi} v_{oi})
 
@@ -710,25 +714,27 @@ These statistics are the moments for wind vector values, where **u** is the E-W
 Vector Anomaly L1 and L2 Values
 -------------------------------
 
-Called "UFABAR", "VFABAR", "UOABAR", "VOABAR", "UVFOABAR", "UVFFABAR", "UVOOABAR" in VAL1L2 output :numref:`table_PS_format_info_VAL1L2`
+Called "UFABAR", "VFABAR", "UOABAR", "VOABAR", "UVFOABAR", "UVFFABAR", and "UVOOABAR" in VAL1L2 output :numref:`table_PS_format_info_VAL1L2`
 
-These statistics require climatological values for the wind vector components, :math:`u_c \text{ and } v_c`. The measures are defined below:
+These statistics require climatological values for the wind vector components, where :math:`{u_c}_f` and :math:`{v_c}_f` are the forecast climatology vectors and :math:`{u_c}_o` and :math:`{v_c}_o` are the observation climatology vectors. The measures are defined below:
 
 .. math:: 
-  \text{UFABAR} = \text{Mean}(u_f - u_c) = \frac{1}{n} \sum_{i=1}^n (u_{fi} - u_c)
+  \text{UFABAR} = \text{Mean}(u_f - {u_c}_f) = \frac{1}{n} \sum_{i=1}^n ({u_f}_i - {{u_c}_f}_i)
 
-  \text{VFBAR} = \text{Mean}(v_f - v_c) = \frac{1}{n} \sum_{i=1}^n (v_{fi} - v_c)
+  \text{VFBAR} = \text{Mean}(v_f - {v_c}_f) = \frac{1}{n} \sum_{i=1}^n ({v_f}_i - {{v_c}_f}_i)
 
-  \text{UOABAR} = \text{Mean}(u_o - u_c) = \frac{1}{n} \sum_{i=1}^n (u_{oi} - u_c)
+  \text{UOABAR} = \text{Mean}(u_o - {u_c}_o) = \frac{1}{n} \sum_{i=1}^n ({u_o}_i - {{u_c}_o}_i)
 
-  \text{VOABAR} = \text{Mean}(v_o - v_c) = \frac{1}{n} \sum_{i=1}^n (v_{oi} - v_c)
+  \text{VOABAR} = \text{Mean}(v_o - {v_c}_o) = \frac{1}{n} \sum_{i=1}^n ({v_o}_i - {{v_c}_o}_i)
 
-  \text{UVFOABAR} &= \text{Mean}[(u_f - u_c)(u_o - u_c) + (v_f - v_c)(v_o - v_c)] \\
-                  &= \frac{1}{n} \sum_{i=1}^n (u_{fi} - u_c) + (u_{oi} - u_c) + (v_{fi} - v_c)(v_{oi} - v_c)
+  \text{UVFOABAR} &= \text{Mean}[(u_f - {u_c}_f)(u_o - {u_c}_o) + (v_f - {v_c}_f)(v_o - {v_c}_o)] \\
+                  &= \frac{1}{n} \sum_{i=1}^n ({u_f}_i - {{u_c}_f}_i) + ({u_o}_i - {{u_c}_o}_i) + ({v_f}_i - {{v_c}_f}_i)({v_o}_i - {{v_c}_o}_i)
 
-  \text{UVFFABAR} = \text{Mean}[(u_f - u_c)^2 + (v_f - v_c)^2] = \frac{1}{n} \sum_{i=1}^n ((u_{fi} - u_c)^2 + (v_{fi} - v_c)^2)
+  \text{UVFFABAR} &= \text{Mean}[(u_f - {u_c}_f)^2 + (v_f - {v_c}_f)^2] \\
+                  &= \frac{1}{n} \sum_{i=1}^n (({u_f}_i - {{u_c}_f}_i)^2 + ({v_f}_i - {{v_c}_f}_i)^2)
 
-  \text{UVOOABAR} = \text{Mean}[(u_o - u_c)^2 + (v_o - v_c)^2] = \frac{1}{n} \sum_{i=1}^n ((u_{oi} - u_c)^2 + (v_{oi} - v_c)^2)
+  \text{UVOOABAR} &= \text{Mean}[(u_o - {u_c}_o)^2 + (v_o - {v_c}_o)^2] \\
+                  &= \frac{1}{n} \sum_{i=1}^n (({u_o}_i - {{u_c}_o}_i)^2 + ({v_o}_i - {{v_c}_o}_i)^2)
 
 Gradient Values
 ---------------

docs/Users_Guide/point-stat.rst

@@ -1204,7 +1204,7 @@ The first set of header columns are common to all of the output files generated
     - Mean(o²)
   * - 31
     - MAE
-    - Mean Absolute Error
+    - Mean(\|f-o\|)
 
 .. _table_PS_format_info_SAL1L2:
 
@@ -1223,25 +1223,25 @@ The first set of header columns are common to all of the output files generated
     - Scalar Anomaly L1L2 line type
   * - 25
     - TOTAL
-    - Total number of matched triplets of forecast (f), observation (o), and climatological value (c)
+    - Total number of matched pairs of forecast (f), observation (o), forecast climatology (cf), and observation climatology (co)
   * - 26
     - FABAR
-    - Mean(f-c)
+    - Mean(f-cf)
   * - 27
     - OABAR
-    - Mean(o-c)
+    - Mean(o-co)
   * - 28
     - FOABAR
-    - Mean((f-c)*(o-c))
+    - Mean((f-cf)*(o-co))
   * - 29
     - FFABAR
-    - Mean((f-c)²)
+    - Mean((f-cf)²)
   * - 30
     - OOABAR
-    - Mean((o-c)²)
+    - Mean((o-co)²)
   * - 31
     - MAE
-    - Mean Absolute Error
+    - Mean(\|(f-cf)-(o-co)\|)
 
 .. _table_PS_format_info_VL1L2:
 
@@ -1318,28 +1318,28 @@ The first set of header columns are common to all of the output files generated
     - Vector Anomaly L1L2 line type
   * - 25
     - TOTAL
-    - Total number of matched triplets of forecast winds (uf, vf), observation winds (uo, vo), and climatological winds (uc, vc)
+    - Total number of matched pairs of forecast winds (uf, vf), observation winds (uo, vo), forecast climatology winds (ucf, vcf), and observation climatology winds (uco, vco)
   * - 26
     - UFABAR
-    - Mean(uf-uc)
+    - Mean(uf-ucf)
   * - 27
     - VFABAR
-    - Mean(vf-vc)
+    - Mean(vf-vcf)
   * - 28
     - UOABAR
-    - Mean(uo-uc)
+    - Mean(uo-uco)
   * - 29
     - VOABAR
-    - Mean(vo-vc)
+    - Mean(vo-vco)
   * - 30
     - UVFOABAR
-    - Mean((uf-uc)*(uo-uc)+(vf-vc)*(vo-vc))
+    - Mean((uf-ucf)*(uo-uco)+(vf-vcf)*(vo-vco))
   * - 31
     - UVFFABAR
-    - Mean((uf-uc)²+(vf-vc)²)
+    - Mean((uf-ucf)²+(vf-vcf)²)
   * - 32
     - UVOOABAR
-    - Mean((uo-uc)²+(vo-vc)²)
+    - Mean((uo-uco)²+(vo-vco)²)
   * - 33
     - FA_SPEED_BAR
     - Mean forecast wind speed anomaly
@@ -1348,7 +1348,7 @@ The first set of header columns are common to all of the output files generated
     - Mean observed wind speed anomaly
   * - 35
     - TOTAL_DIR 
-    - Total number of matched triplets for which the forecast, observation, and climatological wind directions are well-defined (i.e. non-zero vectors)
+    - Total number of matched pairs for which the forecast, observation, forecast climatology, and observation climatology wind directions are well-defined (i.e. non-zero vectors)
   * - 36
     - DIRA_ME
     - Mean wind direction anomaly difference, from -180 to 180 degrees

src/libcode/vx_stat_out/stat_columns.cc

@@ -2927,7 +2927,7 @@ void write_sl1l2_cols(const SL1L2Info &sl1l2_info,
       sl1l2_info.oobar);
 
    at.set_entry(r, c+6,  // MAE
-      sl1l2_info.mae);
+      sl1l2_info.smae);
 
    return;
 }
@@ -2963,7 +2963,7 @@ void write_sal1l2_cols(const SL1L2Info &sl1l2_info,
       sl1l2_info.ooabar);
 
    at.set_entry(r, c+6,  // MAE
-      sl1l2_info.mae);
+      sl1l2_info.samae);
 
    return;
 }

src/libcode/vx_statistics/compute_stats.cc

@@ -101,7 +101,7 @@ void compute_cntinfo(const SL1L2Info &s, bool aflag, CNTInfo &cnt_info) {
    cnt_info.me2.v = cnt_info.me.v * cnt_info.me.v;
 
    // Compute mean absolute error
-   cnt_info.mae.v = s.mae;
+   cnt_info.mae.v = s.smae;
 
    // Compute mean squared error
    cnt_info.mse.v = ffbar + oobar - 2.0*fobar;
@@ -1111,7 +1111,7 @@ void compute_sl1l2_mean(const SL1L2Info *sl1l2_info, int n,
          sl1l2_mean.obar   += sl1l2_info[i].obar;
          sl1l2_mean.ffbar  += sl1l2_info[i].ffbar;
          sl1l2_mean.oobar  += sl1l2_info[i].oobar;
-         sl1l2_mean.mae    += sl1l2_info[i].mae;
+         sl1l2_mean.smae   += sl1l2_info[i].smae;
       }
 
       if(sl1l2_info[i].sacount > 0) {
@@ -1121,6 +1121,7 @@ void compute_sl1l2_mean(const SL1L2Info *sl1l2_info, int n,
          sl1l2_mean.oabar   += sl1l2_info[i].oabar;
          sl1l2_mean.ffabar  += sl1l2_info[i].ffabar;
          sl1l2_mean.ooabar  += sl1l2_info[i].ooabar;
+         sl1l2_mean.samae   += sl1l2_info[i].samae;
       }
    } // end for i
 
@@ -1130,13 +1131,14 @@ void compute_sl1l2_mean(const SL1L2Info *sl1l2_info, int n,
       sl1l2_mean.obar  /= n_sl1l2;
       sl1l2_mean.ffbar /= n_sl1l2;
       sl1l2_mean.oobar /= n_sl1l2;
-      sl1l2_mean.mae   /= n_sl1l2;
+      sl1l2_mean.smae  /= n_sl1l2;
    }
    if(sl1l2_mean.sacount > 0) {
       sl1l2_mean.fabar  /= n_sal1l2;
       sl1l2_mean.oabar  /= n_sal1l2;
       sl1l2_mean.ffabar /= n_sal1l2;
       sl1l2_mean.ooabar /= n_sal1l2;
+      sl1l2_mean.samae  /= n_sal1l2;
    }
 
    return;

src/libcode/vx_statistics/met_stats.cc

@@ -1124,11 +1124,11 @@ SL1L2Info & SL1L2Info::operator+=(const SL1L2Info &c) {
       s_info.ffbar = (ffbar*scount + c.ffbar*c.scount)/s_info.scount;
       s_info.oobar = (oobar*scount + c.oobar*c.scount)/s_info.scount;
 
-      if(is_bad_data(mae) || is_bad_data(c.mae)) {
-         s_info.mae = bad_data_double;
+      if(is_bad_data(smae) || is_bad_data(c.smae)) {
+         s_info.smae = bad_data_double;
       }
       else {
-         s_info.mae = (mae*scount + c.mae*c.scount)/s_info.scount;
+         s_info.smae = (smae*scount + c.smae*c.scount)/s_info.scount;
       }
    }
 
@@ -1141,11 +1141,11 @@ SL1L2Info & SL1L2Info::operator+=(const SL1L2Info &c) {
       s_info.ffabar = (ffabar*sacount + c.ffabar*c.sacount)/s_info.sacount;
       s_info.ooabar = (ooabar*sacount + c.ooabar*c.sacount)/s_info.sacount;
 
-      if(is_bad_data(mae) || is_bad_data(c.mae)) {
-         s_info.mae = bad_data_double;
+      if(is_bad_data(samae) || is_bad_data(c.samae)) {
+         s_info.samae = bad_data_double;
       }
       else {
-         s_info.mae = (mae*sacount + c.mae*c.sacount)/s_info.sacount;
+         s_info.samae = (samae*sacount + c.samae*c.sacount)/s_info.sacount;
       }
    }
 
@@ -1170,15 +1170,15 @@ void SL1L2Info::zero_out() {
    // SL1L2 Quantities
    fbar    = obar   = 0.0;
    fobar   = ffbar  = oobar  = 0.0;
+   smae    = 0.0;
    scount  = 0;
 
    // SAL1L2 Quantities
    fabar   = oabar  = 0.0;
    foabar  = ffabar = ooabar = 0.0;
+   samae   = 0.0;
    sacount = 0;
 
-   mae     = 0.0;
-
    return;
 }
 
@@ -1211,6 +1211,7 @@ void SL1L2Info::assign(const SL1L2Info &c) {
    fobar   = c.fobar;
    ffbar   = c.ffbar;
    oobar   = c.oobar;
+   smae    = c.smae;
    scount  = c.scount;
 
    // SAL1L2 Quantities
@@ -1219,10 +1220,9 @@ void SL1L2Info::assign(const SL1L2Info &c) {
    foabar  = c.foabar;
    ffabar  = c.ffabar;
    ooabar  = c.ooabar;
+   samae   = c.samae;
    sacount = c.sacount;
 
-   mae     = c.mae;
-
    return;
 }
 
@@ -1272,7 +1272,7 @@ void SL1L2Info::set(const PairDataPoint &pd_all) {
       fobar += wgt*f*o;
       ffbar += wgt*f*f;
       oobar += wgt*o*o;
-      mae   += wgt*fabs(f-o);
+      smae  += wgt*fabs(f-o);
       scount++;
 
       // SAL1L2 sums
@@ -1282,6 +1282,7 @@ void SL1L2Info::set(const PairDataPoint &pd_all) {
          foabar += wgt*(f-fc)*(o-oc);
          ffabar += wgt*(f-fc)*(f-fc);
          ooabar += wgt*(o-oc)*(o-oc);
+         samae  += wgt*fabs((f-fc)-(o-oc));
          sacount++;
       }
    }

src/libcode/vx_statistics/met_stats.h

@@ -224,17 +224,16 @@ class SL1L2Info {
       double fbar, obar;
       double fobar;
       double ffbar, oobar;
+      double smae;
       int    scount;
 
       // SAL1L2 Quantities
       double fabar, oabar;
       double foabar;
       double ffabar, ooabar;
+      double samae;
       int    sacount;
 
-      // Mean absolute error
-      double mae;
-
       // Compute sums
       void set(const PairDataPoint &);