Bennett, L., Luker, J., English, C., and Hillier, S. (2016). Stroke survivors’ perspectives on two novel models of inpatient rehabilitation: Seven-day a week individual
therapy or five-day a week circuit class therapy. Disability and Rehabilitation, 38(14),
1397–1406.
Bernhardt, J., Dewey, H., Thrift, A., and Donnan, G. (2004). Inactive
and
alone:
physical activity within the first 14 days of acute stroke unit care. Stroke, 35(4),
1005–1009.
Bernhardt, J., Chan, J., Nicola, I., and Collier, J. M. (2007). Little therapy,
little
physical activity: rehabilitation within the first 14 days of organized stroke unit
care. Journal of Rehabilitation Medicine, 39(1), 43–48.
Bode, R. K., Heinemann, A. W., Semik, P., and Mallinson, T. (2004). Patterns
of
therapy activities across length of stay and impairment levels: peering inside the
“black box” of inpatient stroke rehabilitation. Archives of Physical Medicine and
Rehabilitation, 85(12), 1901–1908.
English, C., Bernhardt, J., and Hillier, S. (2014). Circuit class therapy and 7-day-week
therapy increase physiotherapy time, but not patient activity: early results from the
CIRCIT trial. Stroke, 45(10), 3002–3007.
Jette, D. U., Latham, N. K., Smout, R. J., Gassaway, J., Slavin, M. D., et al. (2005).
Physical therapy interventions for patients with stroke in inpatient rehabilitation
facilities. Physical Therapy, 85(3), 238–248.
Kamo, T., Momosaki, R., Suzuki, K., Asahi, R., Azami, M., Ogihara, H., et al. (2019).
Effectiveness of intensive rehabilitation therapy on functional outcomes after stroke:
Efficacy of categories in physical therapy for improving motor function of patients with stroke
a propensity score analysis based on Japan rehabilitation database. Journal of
Stroke and Cerebrovascular Diseases, 28(9), 2537–2542.
Kaur, G., English, C., and Hillier, S. (2012). How physically active are people with
stroke in physiotherapy sessions aimed at improving motor function? A systematic
review. Stroke Research and Treatment. Article ID 820673.
Kuys, S., Brauer, S., and Ada, L. (2006). Routine physiotherapy does not induce a cardiorespiratory training effect post-stroke, regardless of walking ability. Physiotherapy
Research International, 11(4), 219–227.
Leach, E., Cornwell, P., Fleming, J., and Haines, T. (2010). Patient centered goalsetting in a subacute rehabilitation setting. Disability and Rehabilitation, 32(2),
159–172.
Nagai, S., Sonoda, S., Miyai, I., Kakehi, A., Goto, S., Takayama, Y., Ota, T., et al.
(2011). Relationship between the intensity of stroke rehabilitation and outcome:
A survey conducted by the Kaifukuki Rehabilitation Ward Association in Japan
(second report). Japanese Journal of Comprehensive Rehabilitation Science, 2,
77–81.
Sunnerhagen, K. S., Brown, B., and Kasper, C. E. (2003). Sitting up and transferring
to a chair: two functional tests for patients with stroke. Journal of Rehabilitation
Medicine, 35, 180–183.
Tyson, S. F., and Selley, A. (2004). The development of the stroke physiotherapy intervention recording tool (SPIRIT). Disability and Rehabilitation, 26(20), 1184–1188.
Van Peppen, R. P., Kwakkel, G., Wood-Dauphinee, S., Hendriks, H. J., et al. (2004).
The impact of physical therapy on functional outcomes after stroke: what’s the
evidence? Clinical Rehabilitation, 18(8), 833–862.
Wittwer, J. E., Goldie, P. A., Matyas, T. A., and Galea, M. P. (2000). Quantification
of physiotherapy treatment time in stroke rehabilitation-criterion-related validity.
Australian Journal of Physiotherapy, 46(4), 291–298.
Appendix
Consider causal relationship between Z and Y , where Z is the cause and Y is the effect.
Let X be the third variable.
Definition. X is not a confounder in the relationship of Z → Y if and only if
(i) the relationship of Z → Y does not depend on X = x, and
(ii) X and Y are conditionally independent conditioned on Z, or X and Z are
conditionally independent conditioned on Y .
Proposition 1. Suppose that (X, Y, Z) follows three-dimensional normal distribution. Then ρzx = 0, or ρxy = 0 if X is not a confounder in the Z → Y relationships,
where ρzx and ρxy are Pearson correlation coefficients between X and Z, and X and Y ,
respectively.
Proof. Suppose that X is not a confounder in the Z → Y relationship, then by the
definition the conditional Pearson correlation coefficient between Y and Z conditioned on
X = x, denoted by ρ(y, z|x), does not depend on X = x and X and Y are conditionally
10
Y. Yoshida and T. Yanagawa
independent conditioned on Z, or ρ(y, z|x), does not depend on X = x and X and
Z are conditionally independent conditioned on Y . Since X and Y are conditionally
independent conditioned on Z if and only if ρxy = ρyz ρzx , it follows that
1 − ρ2zx
ρyz − ρxy ρzx
ρ(y, z|x) = √
= ρyz
1 − ρ2yz ρ2zx
(1 − ρ2 )(1 − ρ2 )
xy
zx
Thus, ρ(y, z|x) does not depend on X = x if and only if ρzx = 0. Next, suppose
that ρ(y, z|x), does not depend on X = x and X and Z are conditionally independent
conditioned on Y . Then, similarly as above we may show ρxy = 0. (QED)
The proposition shows that if ρzx ̸= 0 and ρxy ̸= 0, then X is the confounder in the
relationship of Z → Y . Furthermore, we have the following proposition in the context
of regression of Y on Z and X.
Proposition 2. Suppose that (X, Y, Z) follows three-dimensional normal distribution. Then, X and Y are conditionally independent conditioned on Z if and only if
E(Y |z, x) = E(Y |z), where E(Y |z, x) is the conditional expectation of Y given Z = z
and X = x, and E(Y |z) is the conditional expectation of Y given Z = z.
Proof. Let µx , µy and µz be expectations of X, Y and Z, respectively, and σx , σy
and σz be standard deviations of X, Y and Z, Then
σy (ρyz − ρxy ρzx )
σy (ρxy − ρyz ρzx )
(x − µx ) + ( )
(z − µz ),
σx
(1 − ρ2zx )
σz
(1 − ρ2zx )
E(Y |z, x)
µy + (
E(Y |z)
σy
ρyz (z − µz )
σz
Thus, if X and Y are conditionally independent conditioned on Z, that is ρ(x, y|z) =
ρxy − ρzx ρyz = 0, then E(Y |z, x) = E(Y |z). The reverse is trivial. (QED).
Received: February 21, 2023
Revised: March 2, 2023
Accept: March 4, 2023
...